├── cpdetect
├── tests
│ ├── __init__.py
│ ├── files
│ │ ├── data.npy
│ │ └── data2.npz
│ ├── utils.py
│ └── test_cp_detector.py
├── __init__.py
├── utils.py
├── nonlinear_filter.py
└── cp_detector.py
├── attic
├── figure.pdf
├── conv2math.py
├── show_plot.py
├── test_cpd.py
├── README.txt
├── test_2stateweights.py
├── 100pts.txt
├── test_splitter.py
├── data.txt
├── trajectoryGenerator.py
├── oldtype-detectionScript.py
├── poissonDetect.py
├── detectionScript.py
├── old-cpDetect.py
├── old2cpDetect.py
├── cpDetect.py
└── trajectory.dat
├── examples
├── synthetic.pdf
├── true_ts.pickle
├── confusion_matrix.pdf
├── confusion_matrix.png
├── test_refinement.pdf
├── step_synthetic.pickle
├── synthetic_trajs.np.npy
├── Confusion_matrix_filtered.pdf
├── refined_confusion_matrix.png
├── ts_log_odds.csv
├── README.md
├── example_nonfiltered.py
├── example_filtered.py
└── Confusion_matrix.ipynb
├── setup.py
├── README.md
└── LICENSE
/cpdetect/tests/__init__.py:
--------------------------------------------------------------------------------
1 |
--------------------------------------------------------------------------------
/cpdetect/__init__.py:
--------------------------------------------------------------------------------
1 | from cpdetect.cp_detector import Detector as cpDetector
--------------------------------------------------------------------------------
/attic/figure.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/attic/figure.pdf
--------------------------------------------------------------------------------
/examples/synthetic.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/synthetic.pdf
--------------------------------------------------------------------------------
/examples/true_ts.pickle:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/true_ts.pickle
--------------------------------------------------------------------------------
/cpdetect/tests/files/data.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/cpdetect/tests/files/data.npy
--------------------------------------------------------------------------------
/examples/confusion_matrix.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/confusion_matrix.pdf
--------------------------------------------------------------------------------
/examples/confusion_matrix.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/confusion_matrix.png
--------------------------------------------------------------------------------
/examples/test_refinement.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/test_refinement.pdf
--------------------------------------------------------------------------------
/cpdetect/tests/files/data2.npz:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/cpdetect/tests/files/data2.npz
--------------------------------------------------------------------------------
/examples/step_synthetic.pickle:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/step_synthetic.pickle
--------------------------------------------------------------------------------
/examples/synthetic_trajs.np.npy:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/synthetic_trajs.np.npy
--------------------------------------------------------------------------------
/examples/Confusion_matrix_filtered.pdf:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/Confusion_matrix_filtered.pdf
--------------------------------------------------------------------------------
/examples/refined_confusion_matrix.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/choderalab/cpdetect/HEAD/examples/refined_confusion_matrix.png
--------------------------------------------------------------------------------
/attic/conv2math.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 | from cPickle import Unpickler
4 | file=open( "trajectory-1.dat" )
5 | out = open( "data.txt", "w" )
6 | u = Unpickler( file )
7 | data = u.load()
8 | for elem in data : out.write( "%f\n" % elem )
9 | file.close()
10 | out.close()
11 |
--------------------------------------------------------------------------------
/attic/show_plot.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | import sys
5 | import cPickle
6 | import matplotlib.pyplot as plt
7 |
8 | filename = sys.argv[1]
9 | file = open( filename )
10 | u = cPickle.Unpickler( file )
11 | trajectory=u.load()
12 |
13 | plt.plot( trajectory, "o", color=(0,0,0) )
14 | plt.show()
15 |
--------------------------------------------------------------------------------
/attic/test_cpd.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | from cpDetect import *
5 |
6 | data = load_testdata()
7 | #print "data:", data
8 | print "number of points:", len(data)
9 |
10 | mean_var_array=calc_mean_mss( data )
11 |
12 | print numpy.array( mean_var_array )
13 | print "number of points:", len(mean_var_array)
14 |
--------------------------------------------------------------------------------
/setup.py:
--------------------------------------------------------------------------------
1 | from setuptools import setup
2 |
3 | setup(name='cpdetect',
4 | description='Bayesian Change point detection',
5 | url='https://github.com/choderalab/cpdetect',
6 | author='Chaya D. Stern',
7 | packages=['cpdetect', 'cpdetect.tests'],
8 | install_requires=[
9 | 'numpy',
10 | 'pandas',
11 | 'scipy'
12 | ])
--------------------------------------------------------------------------------
/examples/ts_log_odds.csv:
--------------------------------------------------------------------------------
1 | ,ts,log_odds,start_end
2 | 0,500.0,69.0431674762,"(0, 4500)"
3 | 0,1000.0,33.6929122613,"(500, 4500)"
4 | 0,1502.0,63.7269084683,"(1000, 4500)"
5 | 0,1970.0,49.1979964847,"(1502, 4500)"
6 | 0,2511.0,13.7408498616,"(1502, 3000)"
7 | 0,3000.0,1.96064583383,"(1502, 2511)"
8 | 0,3502.0,114.607239634,"(3000, 4500)"
9 | 0,3970.0,2.69324946338,"(3502, 4500)"
10 | 1,490.0,222.665465863,"(0, 1500)"
11 | 1,997.0,0.377356252234,"(0, 997)"
12 |
--------------------------------------------------------------------------------
/attic/README.txt:
--------------------------------------------------------------------------------
1 |
2 | Optimization:
3 |
4 | 1. '../fast-means/maybeMeans.py' seemed faster than '../fast-means/slowerMeans.py.' The latter just called mean() and var()
5 | methods repeatedly, while the former keeps track of stacks of numbers (which data is in which pile) and calculates only
6 | N averages, average squares, and variances. However, when implemented in cpDetect it did not seem that much faster.
7 |
8 | result: move cpDetect.py to old-cpDetect.py
9 |
10 | 2. tabulated gamma.
11 | N with table without table
12 | 1000 2.116 s 3.827 s
13 | 10000 31.329 s 60.942 s
14 |
--------------------------------------------------------------------------------
/attic/test_2stateweights.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | import scipy
5 | from cpDetect import *
6 | import matplotlib.pyplot as plt
7 | from mpmath import log, mpf
8 |
9 | FILE=open( "trajectory-1.dat", "r" )
10 | u = cPickle.Unpickler( FILE )
11 | trajectory = scipy.array( u.load() )
12 |
13 | wts = calc_twostate_weights(trajectory)
14 | for i in range( len(wts) ): print i, wts[i]
15 |
16 | # normalize the weights
17 | tot = mpf( 0.0 )
18 | for elem in wts: tot+=elem
19 | print tot
20 |
21 | logwts = []
22 | for elem in wts : logwts.append( log(elem,10 ))
23 |
24 | for i in range( len(logwts) ): print i, logwts[i]
25 |
26 | plt.plot( logwts )
27 | plt.show()
28 |
29 |
--------------------------------------------------------------------------------
/attic/100pts.txt:
--------------------------------------------------------------------------------
1 |
2 | 1
3 | 4
4 | 1
5 | 1
6 | 1
7 | 1
8 | 2
9 | 0
10 | 0
11 | 2
12 | 13
13 | 11
14 | 14
15 | 12
16 | 7
17 | 11
18 | 17
19 | 6
20 | 9
21 | 15
22 | 9
23 | 16
24 | 10
25 | 8
26 | 12
27 | 16
28 | 9
29 | 10
30 | 15
31 | 10
32 | 10
33 | 9
34 | 12
35 | 6
36 | 12
37 | 7
38 | 6
39 | 8
40 | 13
41 | 11
42 | 10
43 | 9
44 | 8
45 | 6
46 | 10
47 | 10
48 | 10
49 | 12
50 | 6
51 | 13
52 | 6
53 | 13
54 | 13
55 | 13
56 | 7
57 | 12
58 | 11
59 | 9
60 | 7
61 | 5
62 | 5
63 | 5
64 | 5
65 | 3
66 | 4
67 | 4
68 | 5
69 | 4
70 | 3
71 | 4
72 | 8
73 | 2
74 | 2
75 | 5
76 | 4
77 | 3
78 | 8
79 | 4
80 | 5
81 | 5
82 | 1
83 | 5
84 | 3
85 | 3
86 | 4
87 | 7
88 | 4
89 | 7
90 | 6
91 | 7
92 | 7
93 | 3
94 | 1
95 | 6
96 | 2
97 | 4
98 | 3
99 | 8
100 | 3
101 | 2
102 |
--------------------------------------------------------------------------------
/cpdetect/tests/utils.py:
--------------------------------------------------------------------------------
1 | from pkg_resources import resource_filename
2 | import os
3 |
4 |
5 | def get_fn(filename, written=False):
6 | """Get the full path to one of the reference files shipped for testing
7 |
8 | These files are in torsionfit/testing/reference
9 |
10 | :param
11 | name: str
12 | Name of file to load
13 |
14 | :returns
15 | fn : str
16 | full path to file
17 | """
18 | if written:
19 | fn = resource_filename('cpdetect', os.path.join('tests', 'files', 'writes', filename))
20 | else:
21 | fn = resource_filename('cpdetect', os.path.join('tests', 'files', filename))
22 |
23 | #if not os.path.exists(fn):
24 | # raise ValueError('%s does not exist. If you just added it you will have to re install' % fn)
25 |
26 | return fn
27 |
--------------------------------------------------------------------------------
/attic/test_splitter.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | from cpDetect import ChangePointDetector
5 | from numpy import array, ones, column_stack
6 | from random import randint
7 |
8 | NONE = None
9 |
10 | datalen=20
11 | data = ones( datalen )
12 | for i in range( randint(1,10 ) ) :
13 | data[ randint(0, datalen-1) ] = randint(2, 100)
14 | data[17]=101.0
15 | print "data length", len(data)
16 | timepts = array( range(datalen) )
17 |
18 | print "arg max:", data.argmax()
19 |
20 | def function( data ):
21 | npts = len(data)
22 | if npts > 5 :
23 | max=3+data[3:-2].argmax()
24 | min=3+data[3:-2].argmin()
25 | if data[max] > data[min] :
26 | return max
27 | else: return NONE
28 | else: return NONE
29 |
30 | def function2( data ):
31 | # if the 'data' array is not empty, return
32 | if len(data)>0:
33 | amax = data.argmax()
34 | else: return NONE
35 | if data[amax] > 1: return amax
36 | else: return NONE
37 |
38 | cpd= ChangePointDetector( data, function )
39 | cpd.split( 0, len( data ), verbose=True )
40 |
41 | print column_stack( (data, timepts) )
42 | #cpd.changepoints.sort()
43 | print cpd.changepoints
44 |
--------------------------------------------------------------------------------
/attic/data.txt:
--------------------------------------------------------------------------------
1 |
2 | 49.680199
3 | 51.667540
4 | 50.238841
5 | 50.341118
6 | 50.247083
7 | 48.861537
8 | 52.288843
9 | 50.039305
10 | 51.254395
11 | 50.811335
12 | 49.152157
13 | 50.834970
14 | 51.450416
15 | 51.562621
16 | 51.663200
17 | 52.421306
18 | 51.128993
19 | 51.709788
20 | 52.560860
21 | 52.771587
22 | 51.887091
23 | 52.525995
24 | 52.312740
25 | 52.289705
26 | 50.979121
27 | 51.848799
28 | 51.048845
29 | 51.197734
30 | 52.238135
31 | 52.544505
32 | 51.618252
33 | 51.549746
34 | 51.305986
35 | 52.403563
36 | 53.080316
37 | 52.030125
38 | 52.679011
39 | 52.098904
40 | 52.169686
41 | 52.445361
42 | 52.262756
43 | 44.028987
44 | 49.448679
45 | 46.051843
46 | 38.944709
47 | 43.456068
48 | 45.249204
49 | 44.993101
50 | 49.017942
51 | 47.799100
52 | 49.808166
53 | 50.179213
54 | 37.723214
55 | 44.332136
56 | 39.117955
57 | 38.265214
58 | 38.490289
59 | 44.834664
60 | 43.039055
61 | 47.070771
62 | 42.509371
63 | 45.939831
64 | 44.077206
65 | 37.793076
66 | 40.543037
67 | 41.167947
68 | 48.208440
69 | 49.384098
70 | 40.241809
71 | 40.299624
72 | 44.093911
73 | 45.972379
74 | 37.362315
75 | 45.589540
76 | 41.887643
77 | 45.379447
78 | 52.574653
79 | 51.194650
80 |
--------------------------------------------------------------------------------
/cpdetect/utils.py:
--------------------------------------------------------------------------------
1 | """
2 | This file is part of cpdetect (change point detection)
3 |
4 | Author: Chaya D. Stern
5 | Date: 11-30-2016
6 | """
7 |
8 | import logging
9 | import sys
10 |
11 | verbose = False
12 |
13 |
14 | def logger(name='cpDetector', pattern='%(asctime)s %(levelname)s %(name)s: %(message)s',
15 | date_format='%H:%M:%S', handler=logging.StreamHandler(sys.stdout)):
16 | """
17 | Retrieves the logger instance associated to the given name
18 | :param name: The name of the logger instance
19 | :param pattern: The associated pattern
20 | :param date_format: The date format to be used in the pattern
21 | :param handler: The logging handler
22 | :return: The logger
23 | """
24 | _logger = logging.getLogger(name)
25 | _logger.setLevel(log_level(verbose))
26 |
27 | if not _logger.handlers:
28 | formatter = logging.Formatter(pattern, date_format)
29 | handler.setFormatter(formatter)
30 | handler.setLevel(log_level(verbose))
31 | _logger.addHandler(handler)
32 | _logger.propagate = False
33 | return _logger
34 |
35 |
36 | def log_level(verbose=verbose):
37 | if verbose:
38 | return logging.DEBUG
39 | else:
40 | return logging.INFO
41 |
--------------------------------------------------------------------------------
/attic/trajectoryGenerator.py:
--------------------------------------------------------------------------------
1 |
2 | import cPickle
3 | import random
4 |
5 | maxnpts=5000;
6 | nstates = 3 # int( random.uniform( 5, 15) )
7 | davg_max = random.uniform( 5, 15 )
8 | dsig_max = 5 # random.uniform( .1, .3 )
9 |
10 | # initial values
11 | avg=50
12 | sigma=1
13 | ptrans = 0.1 # random.uniform( 0.001, .05 )
14 |
15 | # build the trajectory
16 | npts = 0
17 | state = 0
18 | trajectory=[]
19 | statemod=200
20 | while True :
21 |
22 | sample = random.gauss( avg, sigma )
23 | trajectory.append( sample )
24 | npts += 1
25 |
26 | # quit if the maximum has been reached
27 | if npts >= maxnpts : break
28 | """
29 | # test for a new state
30 | if random.random() < ptrans :
31 | print "!! change point at %d !!" % npts
32 | ptrans = random.uniform( 0.001, .05 )
33 | avg = avg+random.uniform( -davg_max, davg_max )
34 | sigma = abs( sigma+random.uniform( -dsig_max, dsig_max ) )
35 | state += 1
36 |
37 | # test if we need a new state
38 | if state > nstates: break
39 | """
40 | try:
41 | if (npts%statemod) == 0:
42 | print "change of state!"
43 | statemod=statemod+random.randint(-100,100)
44 | avg += 10
45 | sigma += 1
46 | except ZeroDivisionError: pass
47 |
48 |
49 | FILE=open("trajectory-1.dat","w")
50 | p=cPickle.Pickler( FILE )
51 | p.dump( trajectory )
52 |
--------------------------------------------------------------------------------
/attic/oldtype-detectionScript.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | from time import time
5 | from old2cpDetect import *
6 | import matplotlib.pyplot as plt
7 | import scipy
8 | from numpy import ones
9 | from mpmath import mpf, gamma
10 | import cPickle
11 |
12 | FILE=open( "trajectory-1.dat", "r" )
13 | u = cPickle.Unpickler( FILE )
14 | trajectory = scipy.array( u.load() )
15 |
16 | cpd = ChangePointDetector( trajectory,findGaussianChangePoint )
17 |
18 | # the part we want to time
19 | t0 = time()
20 | print "splitting"
21 | cpd.split_init()
22 | t1 = time()
23 | print "took %3.3f seconds" % ( t1-t0 )
24 | cpd.sort()
25 | cpd.showall()
26 | print "found %d change points" % cpd.nchangepoints()
27 |
28 | times = array( range( len(trajectory) ) )
29 |
30 | ymax= trajectory.max() * 1.01
31 | ymin= trajectory.min() * 0.99
32 |
33 | # get the shade scale. Any cp will have log B > 0, so scale to the biggest one
34 | for i in range( cpd.nchangepoints() ):
35 | changepoint = cpd.changepoints[i]
36 | y = ( ymin, ymax )
37 | x = changepoint * ones( len(y) )
38 | logodds = cpd.logodds[changepoint]
39 |
40 | if logodds > 1: color=(0,0,0) # black
41 | elif logodds > 0.477: color=(.5,.5,.5) # dark gray
42 | else : color=(.7,.7,.7) # light gray
43 |
44 | print "point ", changepoint, "log odds", logodds, "color", color
45 | plt.plot( x, y, "-", color=color, linewidth=1.5 )
46 |
47 | plt.plot( times, trajectory, "b-" )
48 | plt.savefig( "figure.eps" )
49 | plt.show()
50 |
--------------------------------------------------------------------------------
/attic/poissonDetect.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | import matplotlib.pyplot as plt
5 | from time import time
6 | from numpy.random import poisson
7 | from numpy import concatenate, ones
8 | from cpDetect import *
9 | from mpmath import gamma
10 |
11 | l1=5 ; l2=8
12 | n1=150 ; n2=100
13 | d1=poisson( l1, n1 )
14 | d2=poisson( l2, n2 )
15 | trajectory = concatenate( (d1,d2,d1,d1,d2) )
16 |
17 | # build factorial table
18 | factorial=[]
19 | maxc = trajectory.sum()+2
20 | for i in range(1,maxc):
21 | factorial.append( gamma(i) )
22 |
23 | cpd = ChangePointDetector( trajectory,findPoissonChangePoint,factorial)
24 |
25 | # the part we want to time
26 | t0 = time()
27 | print "splitting"
28 | cpd.split_init()
29 | t1 = time()
30 | print "took %3.3f seconds" % ( t1-t0 )
31 | cpd.sort()
32 | cpd.showall()
33 | print "found %d change points" % cpd.nchangepoints()
34 |
35 | times = array( range( len(trajectory) ) )
36 |
37 | ymax= trajectory.max() * 1.01
38 | ymin= trajectory.min() * 0.99
39 |
40 | # get the shade scale. Any cp will have log B > 0, so scale to the biggest one
41 | for i in range( cpd.nchangepoints() ):
42 | changepoint = cpd.changepoints[i]
43 | y = ( ymin, ymax )
44 | x = changepoint * ones( len(y) )
45 | logodds = cpd.logodds[changepoint]
46 |
47 | if logodds > 1: color=(0,0,0) # black
48 | elif logodds > 0.477: color=(.5,.5,.5) # dark gray
49 | else : color=(.7,.7,.7) # light gray
50 |
51 | print "point ", changepoint, "log odds", logodds, "color", color
52 | plt.plot( x, y, "-", color=color, linewidth=1.5 )
53 |
54 | plt.plot( times, trajectory, "b-" )
55 | plt.savefig( "figure.eps" )
56 | plt.show()
57 |
--------------------------------------------------------------------------------
/attic/detectionScript.py:
--------------------------------------------------------------------------------
1 |
2 | #!/usr/bin/python
3 |
4 | from time import time
5 | from cpDetect import *
6 | import matplotlib.pyplot as plt
7 | import scipy
8 | from numpy import ones
9 | from mpmath import mpf, gamma
10 | import cPickle
11 |
12 | FILE=open( "trajectory-1.dat", "r" )
13 | u = cPickle.Unpickler( FILE )
14 | trajectory = scipy.array( u.load() )
15 |
16 | # generate gamma function table
17 | t0=time()
18 | print "generating gamma function table"
19 | gammatable=[-99,-99,-99] # first three points don't make sense
20 | for i in range(3,len(trajectory)+1): gammatable.append( gamma( 0.5*i - 1 ) )
21 | t1=time()
22 | print "took %3.3f seconds" % ( t1-t0 )
23 |
24 | cpd = ChangePointDetector( trajectory,findGaussianChangePoint, gammatable )
25 |
26 | # the part we want to time
27 | t0 = time()
28 | print "splitting"
29 | cpd.split_init()
30 | t1 = time()
31 | print "took %3.3f seconds" % ( t1-t0 )
32 | cpd.sort()
33 | cpd.showall()
34 | print "found %d change points" % cpd.nchangepoints()
35 |
36 | times = array( range( len(trajectory) ) )
37 |
38 | ymax= trajectory.max() * 1.01
39 | ymin= trajectory.min() * 0.99
40 |
41 | # get the shade scale. Any cp will have log B > 0, so scale to the biggest one
42 | for i in range( cpd.nchangepoints() ):
43 | changepoint = cpd.changepoints[i]
44 | y = ( ymin, ymax )
45 | x = changepoint * ones( len(y) )
46 | logodds = cpd.logodds[changepoint]
47 |
48 | if logodds > 1: color=(0,0,0) # black
49 | elif logodds > 0.477: color=(.5,.5,.5) # dark gray
50 | else : color=(.7,.7,.7) # light gray
51 |
52 | print "point ", changepoint, "log odds", logodds, "color", color
53 | plt.plot( x, y, "-", color=color, linewidth=1.5 )
54 |
55 | plt.plot( times, trajectory, "b-" )
56 | plt.savefig( "figure.eps" )
57 | plt.show()
58 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | cpDetect
2 | ========
3 |
4 | Bayesian change point detection
5 |
6 | Installation
7 | -------------
8 | Install directly from source directory.
9 | `python setup.py install`
10 |
11 | Requirement
12 | -------------
13 | * python 2.7
14 | * NumPy
15 | * Pandas
16 | * SciPy
17 |
18 | Usage
19 | -----
20 |
21 | To run `cpDetect`, the timeseries data needs to be a list of 1-D numpy arrays. They do not have to be of the same size
22 | First, instantiate the detector. Choose the underlying distribution (normal or log normal) and the log odds threshold
23 | (default is 0).
24 |
25 | ```
26 | detector = cpDetector(trajs, distribution='log_normal', log_odds_threshold=0)
27 | detector.detect_cp()
28 | ```
29 |
30 | `cpDetect` can sometimes miss fast transitions. If you find that this is the case for your data, you can try the refinement step
31 | (see `refinement.ipynb` in `examples/` for an illustration how this step works.
32 |
33 | ```
34 | detector.refinement(threshold=-2, reject_window=20, split_windor=50)
35 | ```
36 |
37 | The results of the refinement step are stored in the `.refined_change_point` dictionary. You can also regenerate the
38 | step function:
39 |
40 | ```
41 | detector.regenerate_step_function()
42 | ```
43 |
44 | Save the change points, log odds and the start, end for each segment:
45 |
46 | ```
47 | detector.to_csv('filename.csv')
48 | ```
49 |
50 | You can save the step function to a pandas data frame:
51 |
52 | ```
53 | df = pd.DataFreame.from_dict(detector.step_function, orient='index')
54 | df.to_csv('step_function.csv')
55 | ```
56 |
57 | See `examples/` for confusion matrices and more on the refinement step.
58 |
59 | Filtering
60 | ---------
61 |
62 | `nonlinear_filter.py` implements the non-linear filter from Chung and Kennedy [DOI](https://www.ncbi.nlm.nih.gov/pubmed/1795554)
63 |
64 | Reference
65 | ---------
66 | * Ensign DL and Pande VS. Bayesian Detection of Intensity Changes in Single Molecule and Molecular Dynamics Trajectories.
67 | J. Phys. Chem B 114:280 (2010) [DOI](http://pubs.acs.org/doi/abs/10.1021/jp906786b)
68 |
69 |
70 |
71 |
72 |
73 |
74 |
75 |
--------------------------------------------------------------------------------
/examples/README.md:
--------------------------------------------------------------------------------
1 | Examples
2 | ========
3 |
4 | All examples are on randomly generated synthetic data.
5 |
6 | Manifest
7 | --------
8 | * **Simple example:**
9 | * `example.ipynb` - simple example on 2 trajectories
10 | * `ts_log_odss.csv` - saved change points from simple example
11 |
12 | * **Synthtic Data:**
13 | * `synthetic_data.ipynb` - notebook that generated synthetic trajectories
14 | * `synthetic.pdf` - plots of all synthetic trajectories and true step functions (in red).
15 | * `synthetic_trajs.np.npy` - synthetic trajectories.
16 | * `true_ts.pickle` - pickled dictionary of true change points for synthetic trajectories.
17 | * `step_synthetic.pickle` - true step function for synthetic trajectories.
18 |
19 | * **Confusion matrices:**
20 | * `example_nonfiltered.py` - run cpDetect with several different thresholds
21 | to calculate confusion matrices for each. Data was not filtered.
22 | * `example_filtered.py` - run cpDetect on filtered data with several different
23 | thresholds.
24 | * `confusion_matrix.ipynb` - ipython notebook that calculates the confusion matrix
25 | for each threshold. Unfiltered data
26 | * `confusion_matrix_filtered.ipynb` - ipython notebook that calculates the confusion
27 | matrix for each threshold. Data is filtered.
28 | * `confusion_matrix.pdf` - confusion matrices for all 10 runs. Unfiltered data
29 | * `confusion_matrix_filtered.pdf` - confusion matrices for all 10 runs. Filtered data
30 |
31 | While cpDetect on filtered data misses less change points, it also founds many
32 | false positives. Use a higher threshold when data is filtered.
33 |
34 | * **Clean up step (refinement):**
35 | * `refinement.ipynb` - ipython notebook illustrating refinement step
36 | * `confusion_matrix.png` - confusion matrix before refinement
37 | * `refined_confusion_matrix.png` - confusion matrix after refinement
38 | * `test_refinement.pdf` - results from refinement step on synthetic trajectories.
39 | red - true step function; black - predicted step function from initial cpDetect run;
40 | green lines - change points found with refinement step.
41 |
42 |
--------------------------------------------------------------------------------
/examples/example_nonfiltered.py:
--------------------------------------------------------------------------------
1 | import numpy as np
2 | from cpdetect import cpDetector, nonlinear_filter
3 | import pandas as pd
4 | from tqdm import *
5 | try:
6 | import cPickle as pickle
7 | except:
8 | import pickle
9 | import matplotlib.pyplot as plt
10 | import matplotlib as mpl
11 | from matplotlib.backends.backend_pdf import PdfPages
12 |
13 |
14 | # Load synthetic trajectories
15 | trajs = np.load('synthetic_trajs.np.npy')
16 | true_step = pickle.load(open('step_synthetic.pickle', 'rb'))
17 | trajs = trajs.tolist()
18 |
19 | # Run them through cpdetect for thresholds (-10, 0)
20 | for i in range(10):
21 | detector = cpDetector(trajs, distribution='log_normal', log_odds_threshold=-i)
22 | detector.detect_cp()
23 | # pickle detector
24 | pickle.dump(detector, open('detector_{}.pickle'.format(str(i)), 'wb'))
25 | # save steps
26 | df = pd.DataFrame.from_dict(detector.step_function, orient='index')
27 | df.to_csv('step_function_{}.csv'.format(str(i)))
28 | detector.to_csv('ts_log_odds_{}.csv'.format(str(i)))
29 |
30 | # Plot
31 | filename = 'synthetic_{}.pdf'.format(str(i))
32 | fontsize = 6
33 | x_spacing = 1000
34 | time_res = 1.0
35 | chunk = len(trajs)/4
36 | with PdfPages(filename) as pdf:
37 | for i in tqdm(range(int(chunk))):
38 | fig = plt.figure()
39 | for j in range(4):
40 | ax = fig.add_subplot(2, 2, j+1)
41 | ax.plot(trajs[4*i + j], alpha=0.6)
42 | ax.plot(true_step['traj_{}'.format(4*i +j)], color='red', linewidth=1.0)
43 | ax.plot(detector.step_function['traj_{}'.format(str(4*i+j))], 'black',
44 | linewidth=1.0)
45 | if j in (2,3):
46 | #plt.xlim([0, 3580])
47 | #ax.xaxis.limit_range_for_scale(0, 3580)
48 | #ax.xaxis.set_ticks([k*time_res*(3580/4) for k in range(5)])
49 | ax.xaxis.set_label_text('Time (seconds)')
50 | else:
51 | #plt.xlim([0, 3580])
52 |
53 | ax.xaxis.set_ticks([k*time_res*(3580/4) for k in range(5)])
54 | #ax.xaxis.set_ticks([])
55 | pdf.savefig(bbox_inches='tight')
56 | plt.close()
57 |
58 |
59 |
--------------------------------------------------------------------------------
/examples/example_filtered.py:
--------------------------------------------------------------------------------
1 | import numpy as np
2 | from cpdetect import cpDetector, nonlinear_filter
3 | import pandas as pd
4 | from tqdm import *
5 | try:
6 | import cPickle as pickle
7 | except:
8 | import pickle
9 | import matplotlib.pyplot as plt
10 | import matplotlib as mpl
11 | from matplotlib.backends.backend_pdf import PdfPages
12 |
13 |
14 | # Load synthetic trajectories
15 | trajs = np.load('synthetic_trajs.np.npy')
16 | true_step = pickle.load(open('step_synthetic.pickle', 'rb'))
17 |
18 | # run through filter
19 | windows = [2, 4, 6, 8, 16]
20 | M = 12
21 | p = 30
22 | filtered_trajs = []
23 | for traj in trajs:
24 | filtered_trajs.append(nonlinear_filter.nfl_filter(traj, windows, M, p))
25 |
26 | # save filtered trajs
27 | np.save(file='filtered_trajs', arr=filtered_trajs)
28 |
29 | # Run them through cpdetect for thresholds (-10, 0)
30 | for i in range(11):
31 | detector = cpDetector(filtered_trajs, distribution='log_normal', log_odds_threshold=-i)
32 | detector.detect_cp()
33 | # pickle detector
34 | pickle.dump(detector, open('filtered_detector_{}.pickle'.format(str(i)), 'wb'))
35 | # save steps
36 | df = pd.DataFrame.from_dict(detector.step_function, orient='index')
37 | df.to_csv('filtered_step_function_{}.csv'.format(str(i)))
38 | detector.to_csv('ts_log_odds_filtered_{}.csv'.format(str(i)))
39 |
40 | # Plot
41 | filename = 'synthetic_filtered_{}.pdf'.format(str(i))
42 | fontsize = 6
43 | x_spacing = 1000
44 | time_res = 1.0
45 | chunk = len(trajs)/4
46 | with PdfPages(filename) as pdf:
47 | for i in tqdm(range(int(chunk))):
48 | fig = plt.figure()
49 | for j in range(4):
50 | ax = fig.add_subplot(2, 2, j+1)
51 | ax.plot(trajs[4*i + j], alpha=0.6)
52 | ax.plot(true_step['traj_{}'.format(4*i +j)], color='red', linewidth=1.0)
53 | ax.plot(detector.step_function['traj_{}'.format(str(4*i+j))], 'black',
54 | linewidth=1.0)
55 | if j in (2,3):
56 | #plt.xlim([0, 3580])
57 | #ax.xaxis.limit_range_for_scale(0, 3580)
58 | #ax.xaxis.set_ticks([k*time_res*(3580/4) for k in range(5)])
59 | ax.xaxis.set_label_text('Time (seconds)')
60 | else:
61 | #plt.xlim([0, 3580])
62 |
63 | ax.xaxis.set_ticks([k*time_res*(3580/4) for k in range(5)])
64 | #ax.xaxis.set_ticks([])
65 | pdf.savefig(bbox_inches='tight')
66 | plt.close()
67 |
68 |
69 |
--------------------------------------------------------------------------------
/cpdetect/nonlinear_filter.py:
--------------------------------------------------------------------------------
1 | """
2 | Non-linear filter adapted from Chung and Kennedy
3 | """
4 |
5 | import numpy as np
6 |
7 |
8 | def nfl_filter(traj, windows, M, p):
9 | """
10 |
11 | Parameters
12 | ----------
13 | traj: numpy 1-D array
14 | trajectory
15 | windows: list of ints
16 | lenghts of windows to calculate forward and reverse predictors
17 | M: int
18 | size of window over which the predictors are to be compared
19 | p: int
20 | weighting factor
21 |
22 |
23 | Returns
24 | -------
25 | filtered_traj: numpy 1-D arraay
26 | NLF filtered trajectory
27 | """
28 | start_index = max(windows) + M
29 | filtered_traj = np.zeros(len(traj)-2*start_index)
30 | I_forward = np.zeros((len(windows), len(traj) - 2*start_index))
31 | I_reverse = np.zeros((len(windows), len(traj) - 2*start_index))
32 | for i in range(I_forward.shape[-1]):
33 | for j, k in enumerate(windows):
34 | I_forward[j,i] = forward_predictor(traj, k, i+start_index-1)
35 | I_reverse[j,i] = reverse_predictor(traj, k, i+start_index-1)
36 | for i in range(start_index, len(traj) - start_index):
37 | f_k = np.zeros(len(windows))
38 | b_k = np.zeros(len(windows))
39 | for j, k in enumerate(windows):
40 | for m in range(M-1):
41 | f_k[j] += (traj[i-m] - I_forward[j, i-m-start_index])**2
42 | b_k[j] += (traj[i-m] - I_reverse[j, i-m-start_index])**2
43 | f_k[j] = f_k[j]**-p
44 | b_k[j] = b_k[j]**-p
45 | c = f_k.sum() + b_k.sum()
46 |
47 | for k in range(len(windows)):
48 | filtered_traj[i-start_index] += f_k[k]*I_forward[k, i-start_index] + \
49 | b_k[k]*I_reverse[k, i-start_index]
50 | filtered_traj[i-start_index] /= c
51 | return filtered_traj
52 |
53 |
54 | def forward_predictor(traj, N, i):
55 | """
56 |
57 | Parameters
58 | ----------
59 | traj: numpy array
60 | trajectory
61 | N: int
62 | size of window
63 | i: int
64 | index
65 |
66 | Returns
67 | -------
68 | average intensity of forward window
69 | """
70 | return sum(traj[i-N:i-1])/N
71 |
72 |
73 | def reverse_predictor(traj, N, i):
74 | """
75 |
76 | Parameters
77 | ----------
78 | traj: numpy array
79 | trajectory
80 | N: int
81 | size of window
82 | i: int
83 | index
84 |
85 | Returns
86 | -------
87 | average intensity of reverse window
88 | """
89 | return sum(traj[i+1:i+N])/N
90 |
--------------------------------------------------------------------------------
/cpdetect/tests/test_cp_detector.py:
--------------------------------------------------------------------------------
1 | """ Test Change Point detector """
2 |
3 | from cpdetect import cpDetector
4 | from cpdetect.cp_detector import (Normal, LogNormal)
5 | from cpdetect.tests.utils import get_fn
6 | import numpy as np
7 | import unittest
8 | import pandas as pd
9 | from scipy.special import gammaln
10 |
11 | data = np.load(get_fn('data.npy'))
12 | data_2 = np.load(get_fn('data2.npz'))
13 |
14 | # Convert to cpDetect format
15 | data_2 = [data_2[i] for i in data_2.files]
16 | detector = cpDetector(data_2, distribution='log_normal', log_odds_threshold=-10)
17 |
18 |
19 | class TestCpDetect(unittest.TestCase):
20 |
21 | def test_lognormal(self):
22 | """ Test Log-normal mean and variance """
23 |
24 | mean, var = LogNormal.mean_var(data)
25 | self.assertEqual(mean, 1.0081320131891722)
26 | self.assertEqual(var, 0.010999447786412363)
27 |
28 | def test_normal(self):
29 | """Test Normal mean and variance"""
30 |
31 | mean, var = Normal.mean_var(data)
32 | self.assertEqual(mean, 2.7556468814327055)
33 | self.assertEqual(var, 0.084910408984515948)
34 |
35 | def test_detector_init(self):
36 | """ Test cp detector initiation """
37 |
38 | self.assertEqual(detector.distribution, 'log_normal')
39 | self.assertEqual(detector.nobservations, 2)
40 | self.assertEqual(detector.observation_lengths, [2500, 3000])
41 | self.assertEqual(detector.threshold, -10)
42 |
43 | def test_log_gamma(self):
44 | """ Test log gamma """
45 | gammas = [-99, -99, -99]
46 | for i in range(3, max(detector.observation_lengths) + 1):
47 | gammas.append(gammaln(0.5*i - 1))
48 |
49 | self.assertEqual(gammas, detector.loggamma)
50 |
51 | def test_split(self):
52 | """ Test split """
53 |
54 | def test_log_odds(self):
55 | """ test log odds calculation and finding ts """
56 | p, t, log_odds = detector._normal_lognormal_bf(data)
57 | self.assertEqual(t, 1436)
58 | self.assertAlmostEqual(log_odds, 5.1818026662176635, 2)
59 |
60 | def test_cpdetect(self):
61 | """ Test the detector """
62 |
63 | detector.detect_cp()
64 |
65 | data = {'ts':[1500, 1019], 'log_odds': [21.322622, -6.867833],
66 | 'start_end': [(0, 2500), (0, 1500)]}
67 | df = pd.DataFrame(data, columns=['ts', 'log_odds', 'start_end'])
68 | self.assertTrue(df['ts'].equals(detector.change_points['traj_0']['ts']))
69 | self.assertTrue(df['start_end'].equals(detector.change_points['traj_0']['start_end']))
70 | self.assertAlmostEqual(df['log_odds'][0], detector.change_points['traj_0']['log_odds'][0], 1)
71 | self.assertTrue(len(detector.change_points['traj_0']), 2)
72 | self.assertTrue(len(detector.change_points['traj_1']), 3)
73 |
74 |
75 |
76 |
--------------------------------------------------------------------------------
/attic/old-cpDetect.py:
--------------------------------------------------------------------------------
1 |
2 | # function to quickly calculate the means and means sums of squares of
3 | # all the partitions of a set of points
4 |
5 | import cPickle
6 | import numpy
7 | from math import pi
8 | from scipy import array
9 | from mpmath import log, gamma, mpf # arbitrary float precision!
10 |
11 | def load_testdata( filename = "trajectory.dat" ):
12 | FILE = open( filename )
13 | u = cPickle.Unpickler( FILE )
14 | data = u.load()
15 | FILE.close()
16 | return data
17 |
18 | # for Gaussian noise only. For this reason, the first three points and the last two points
19 | # cannot correspond to a change point. Thus, we return an array of length (npts-5).
20 | def calc_mean_mss( data ):
21 | #data = numpy.array( data, "float64" )
22 | npts = len( data )
23 |
24 | mean_var_array=[]
25 |
26 | points = range( 3, npts-2 )
27 | for i in points :
28 | dataA = data[ 0:i ]
29 | dataB = data[ i: ]
30 |
31 | lenA = len( dataA )
32 | lenB = len( dataB )
33 |
34 | dataA2 = dataA**2
35 | dataB2 = dataB**2
36 |
37 | mean_dataA = dataA.mean()
38 | mean_dataB = dataB.mean()
39 |
40 | mean_dataA2 = (dataA2).mean()
41 | mean_dataB2 = (dataB2).mean()
42 |
43 | mean2_dataA = mean_dataA**2
44 | mean2_dataB = mean_dataB**2
45 |
46 | varA = mean_dataA2 - mean2_dataA
47 | varB = mean_dataB2 - mean2_dataB
48 |
49 | mean_var_array.append( (lenA, mean2_dataA, varA, lenB, mean2_dataB, varB) )
50 | return mean_var_array
51 |
52 | # uses calc_mean_mss() to compute the relative weights of the switch time
53 | def calc_twostate_weights( data ):
54 | weights=[0,0,0] # the change cannot have occurred in the last 3 points
55 | means_mss=calc_mean_mss( data )
56 |
57 | i=0
58 | try:
59 | for nA, mean2A, varA, nB, mean2B, varB in means_mss :
60 | #print "computing for data", nA, mean2A, varA, nB, mean2B, varB
61 | numf1 = calc_alpha( nA, mean2A, varA )
62 | numf2 = calc_alpha( nB, mean2B, varB )
63 | denom = (varA + varB) * (mean2A*mean2B)
64 | weights.append( (numf1*numf2)/denom)
65 | i += 1
66 | except:
67 | print "failed at data", i # means_mss[i]
68 | print "---"
69 | print means_mss
70 | print "---"
71 | raise
72 |
73 | weights.extend( [0,0] ) # the change cannot have occurred at the last 2 points
74 | return array( weights )
75 |
76 | def calc_alpha( N, x2, s2 ):
77 | first = mpf(N)**(-N/2.0 + 1.0/2.0)
78 | second = mpf(s2)**(-N/2.0 + 1.0 )
79 | third = gamma( mpf(N)/2.0 - 1.0 )
80 | return first*second*third
81 |
82 | def findGaussianChangePoint( data ):
83 |
84 | # the denominator. This is the easy part.
85 | N = len( data )
86 |
87 | if N<6 : return None # can't find a cp in data this small
88 |
89 | s2 = mpf(data.var())
90 | gpart = gamma( mpf(N)/2.0 - 1 )
91 | denom = (pi**1.5) * mpf((N*s2))**( -N/2.0 + 0.5 ) * gpart
92 |
93 | # the numerator. A little trickier.
94 | # calc_twostate_weights() already deals with ts<3 and ts>N-2.
95 | weights=calc_twostate_weights( data )
96 | if weights is None: return None
97 |
98 | num = 2.0**2.5 * abs(data.mean()) * weights.mean()
99 |
100 | lognum=log(num,10)
101 | logdenom=log(denom,10)
102 | logodds = lognum-logdenom
103 |
104 | print "num:", num, "log num:", lognum, "| denom:", denom, "log denom:", logdenom, "|| log odds:", logodds
105 |
106 | # If there is a change point, then logodds will be greater than 0
107 | if logodds < 0 :
108 | return None
109 |
110 | return ( weights.argmax(), logodds )
111 |
112 | class ChangePointDetector:
113 | def __init__( self, data, function ):
114 | self.data = data
115 | self.datalen = len( self.data )
116 | self.function = function
117 | self.changepoints = []
118 | self.logodds = {}
119 | self.niter = 0
120 | self.maxiter = 1000000 # just in case
121 |
122 | def nchangepoints( self ):
123 | return len( self.changepoints )
124 |
125 | def split_init( self, verbose=False ):
126 | self.split( 0, self.datalen, verbose )
127 |
128 | def split( self, start, end, verbose=False ):
129 | if self.niter > self.maxiter :
130 | print "Change point detection error: number of iterations exceeded"
131 | print "If this is the right result, you may need to increase"
132 | print "ChangePointDetector.maxiter (currently %d)" % self.maxiter
133 | return
134 | self.niter += 1
135 | if verbose:
136 | print "\nIteration %d" % self.niter
137 | print "Trying to split the segment:", self.data[start:end], "(data from %d to %d)" % ( start, end)
138 | print self.data[start:end]
139 |
140 | # try to find a change point in the data segment
141 | try:
142 | result = self.function( self.data[ start: end ] )
143 | except TypeError:
144 | print "trying to test data from %d to %d failed" % ( start,end )
145 | #print self.data
146 | raise
147 |
148 | # otherwise, store the cp and call self.split on the two ends
149 | if result is not None :
150 | try: # fails if only one value is returned
151 | logodds = result[1]
152 | self.logodds[ start+result[0] ] = logodds
153 | result = start+result[0]
154 | except TypeError: # must mean it's one number?
155 | result += start
156 |
157 | if verbose: print "!! change point detected at %d !!" % result
158 | self.changepoints.append( result )
159 | self.split( start, result, verbose )
160 | self.split( result+1, end, verbose )
161 |
162 | def sort( self ): self.changepoints.sort()
163 |
164 | # display the change points
165 | def show( self ): print self.changepoints
166 |
167 | # show the change points along with the log odds
168 | def showall( self ):
169 | for i in range( len( self.changepoints ) ) :
170 | changepoint = self.changepoints[i]
171 | try:
172 | logodds = self.logodds[ changepoint ]
173 | except KeyError:
174 | logodds = None
175 | print "%d (%f)" % ( changepoint, logodds )
176 |
177 | def largest_logodds( self ):
178 | return array( self.logodds.values() ).max()
179 |
--------------------------------------------------------------------------------
/attic/old2cpDetect.py:
--------------------------------------------------------------------------------
1 |
2 | # function to quickly calculate the means and means sums of squares of
3 | # all the partitions of a set of points
4 |
5 | import cPickle
6 | import numpy
7 | from math import pi
8 | from scipy import array
9 | from mpmath import log, gamma, mpf # arbitrary float precision!
10 | import sys
11 |
12 | def load_testdata( filename = "trajectory.dat" ):
13 | FILE = open( filename )
14 | u = cPickle.Unpickler( FILE )
15 | data = u.load()
16 | FILE.close()
17 | return data
18 |
19 | # for Gaussian noise only. For this reason, the first three points and the last two points
20 | # cannot correspond to a change point. Thus, we return an array of length (npts-5).
21 | def calc_mean_mss( data ):
22 | #data = numpy.array( data, "float64" )
23 | npts = len( data )
24 |
25 | #initialize
26 | data2=data**2
27 | dataA=data[0:3]
28 | dataA2=data2[0:3]
29 | NA = len(dataA)
30 | dataB=data[3:]
31 | dataB2=data2[3:]
32 | NB = len(dataB)
33 |
34 | sumA=dataA.sum() ; sumsqA=dataA2.sum()
35 | sumB=dataB.sum() ; sumsqB=dataB2.sum()
36 |
37 | mean_var_array=[]
38 |
39 | # first data point
40 | meanA=sumA/NA
41 | meanB=sumB/NB
42 | meansumsqA = sumsqA/NA
43 | meansumsqB = sumsqB/NB
44 | meanA2 = meanA**2
45 | meanB2 = meanB**2
46 | sA2=meansumsqA-meanA2
47 | sB2=meansumsqB-meanB2
48 | mean_var_array.append( (3, meanA2, sA2, npts-3, meanB2, sB2 ) )
49 |
50 | for i in range( 3, npts-3 ):
51 | NA += 1 ; NB -= 1
52 | next = data[i]
53 | sumA += next ; sumB -= next
54 | nextsq = data2[i]
55 | sumsqA += nextsq; sumsqB -= nextsq
56 |
57 | meanA=sumA/NA
58 | meanB=sumB/NB
59 | meansumsqA=sumsqA/NA
60 | meansumsqB=sumsqB/NB
61 | meanA2=meanA**2
62 | meanB2=meanB**2
63 | sA2=meansumsqA-meanA2
64 | sB2=meansumsqB-meanB2
65 | mean_var_array.append( (NA, meanA2, sA2, NB, meanB2, sB2) )
66 |
67 | return mean_var_array
68 |
69 | # uses calc_mean_mss() to compute the relative weights of the switch time
70 | def calc_twostate_weights( data ):
71 | weights=[0,0,0] # the change cannot have occurred in the last 3 points
72 | means_mss=calc_mean_mss( data )
73 |
74 | i=0
75 | try:
76 | for nA, mean2A, varA, nB, mean2B, varB in means_mss :
77 | #print "computing for data", nA, mean2A, varA, nB, mean2B, varB
78 | numf1 = calc_alpha( nA, mean2A, varA )
79 | numf2 = calc_alpha( nB, mean2B, varB )
80 | denom = (varA + varB) * (mean2A*mean2B)
81 | weights.append( (numf1*numf2)/denom)
82 | i += 1
83 | except:
84 | print "failed at data", i # means_mss[i]
85 | print "---"
86 | #print means_mss
87 | print "---"
88 | raise
89 |
90 | weights.extend( [0,0] ) # the change cannot have occurred at the last 2 points
91 | return array( weights )
92 |
93 | def calc_alpha( N, x2, s2 ):
94 | first = mpf(N)**(-N/2.0 + 1.0/2.0)
95 | second = mpf(s2)**(-N/2.0 + 1.0 )
96 | third = gamma( mpf(N)/2.0 - 1.0 )
97 | return first*second*third
98 |
99 | def findGaussianChangePoint( data ):
100 |
101 | # the denominator. This is the easy part.
102 | N = len( data )
103 |
104 | if N<6 : return None # can't find a cp in data this small
105 |
106 | # set up gamma function table
107 | #for i in range(N):
108 |
109 |
110 | s2 = mpf(data.var())
111 | gpart = gamma( mpf(N)/2.0 - 1 )
112 | denom = (pi**1.5) * mpf((N*s2))**( -N/2.0 + 0.5 ) * gpart
113 |
114 | # the numerator. A little trickier.
115 | # calc_twostate_weights() already deals with ts<3 and ts>N-2.
116 | weights=calc_twostate_weights( data )
117 | if weights is None: return None
118 |
119 | num = 2.0**2.5 * abs(data.mean()) * weights.mean()
120 |
121 | logodds = log( num ) - log( denom )
122 |
123 | print "num:", num, "log num:", log(num), "| denom:", denom, "log denom:", log(denom), "|| log odds:", logodds
124 |
125 | # If there is a change point, then logodds will be greater than 0
126 | if logodds < 0 :
127 | return None
128 |
129 | return ( weights.argmax(), logodds )
130 |
131 | class ChangePointDetector:
132 | def __init__( self, data, function ):
133 | self.data = data
134 | self.datalen = len( self.data )
135 | self.function = function
136 | self.changepoints = []
137 | self.logodds = {}
138 | self.niter = 0
139 | self.maxiter = 1000000 # just in case
140 |
141 | def nchangepoints( self ):
142 | return len( self.changepoints )
143 |
144 | def split_init( self, verbose=False ):
145 | self.split( 0, self.datalen, verbose )
146 |
147 | def split( self, start, end, verbose=False ):
148 | if self.niter > self.maxiter :
149 | print "Change point detection error: number of iterations exceeded"
150 | print "If this is the right result, you may need to increase"
151 | print "ChangePointDetector.maxiter (currently %d)" % self.maxiter
152 | return
153 | self.niter += 1
154 | if verbose:
155 | print "\nIteration %d" % self.niter
156 | print "Trying to split the segment:", self.data[start:end], "(data from %d to %d)" % ( start, end)
157 | print self.data[start:end]
158 |
159 | # try to find a change point in the data segment
160 | try:
161 | result = self.function( self.data[ start: end ] )
162 | except TypeError:
163 | print "trying to test data from %d to %d failed" % ( start,end )
164 | #print self.data
165 | raise
166 |
167 | # otherwise, store the cp and call self.split on the two ends
168 | if result is not None :
169 | try: # fails if only one value is returned
170 | logodds = result[1]
171 | self.logodds[ start+result[0] ] = logodds
172 | result = start+result[0]
173 | except TypeError: # must mean it's one number?
174 | result += start
175 |
176 | if verbose: print "!! change point detected at %d !!" % result
177 | self.changepoints.append( result )
178 | self.split( start, result, verbose )
179 | self.split( result+1, end, verbose )
180 |
181 | def sort( self ): self.changepoints.sort()
182 |
183 | # display the change points
184 | def show( self ): print self.changepoints
185 |
186 | # show the change points along with the log odds
187 | def showall( self ):
188 | for i in range( len( self.changepoints ) ) :
189 | changepoint = self.changepoints[i]
190 | try:
191 | logodds = self.logodds[ changepoint ]
192 | except KeyError:
193 | logodds = None
194 | print "%d (%f)" % ( changepoint, logodds )
195 |
196 | def largest_logodds( self ):
197 | return array( self.logodds.values() ).max()
198 |
--------------------------------------------------------------------------------
/attic/cpDetect.py:
--------------------------------------------------------------------------------
1 |
2 | # function to quickly calculate the means and means sums of squares of
3 | # all the partitions of a set of points
4 |
5 | import cPickle
6 | import numpy
7 | from math import pi
8 | from scipy import array, zeros
9 | from mpmath import log, gamma, mpf # arbitrary float precision!
10 | import sys
11 |
12 | inv_log10=1.0/log(10)
13 |
14 | def load_testdata( filename = "trajectory.dat" ):
15 | FILE = open( filename )
16 | u = cPickle.Unpickler( FILE )
17 | data = u.load()
18 | FILE.close()
19 | return data
20 |
21 | def findPoissonChangePoint( data, factorial ):
22 | # data is a list of counts in each time period, uniformly spaced
23 |
24 | # the denominator (including both P(D|H1) and constant parts of P(D|H2) )
25 | C = data.sum()
26 | N = mpf(len(data))
27 | denominator = factorial[C-1] * pi / ( 2 * N**C )
28 |
29 | # the numerator (trickier)
30 | # this needs to be averaged over the possible change points
31 | weights = zeros(N,dtype=object)
32 | CA = 0
33 | CB = C
34 | for i in range(1,N) :
35 | # points up through i are in data set A; the rest are in B
36 | datapoint = data[i-1]
37 | NA = mpf(i) ; CA += datapoint
38 | NB = mpf(N-i) ; CB -= datapoint
39 |
40 | fraction_num = factorial[CA] * factorial[CB]
41 | fraction_den = NA**(CA+1) * NB**(CB+1) * ( (CA/NA)**2 + (CB/NB)**2 )
42 | #weights.append( fraction_num/fraction_den )
43 | weights[i-1] = mpf(fraction_num)/fraction_den
44 |
45 | numerator = weights.mean()
46 | lognum= inv_log10 * log( numerator )
47 | logden= inv_log10 * log( denominator )
48 | logodds = lognum - logden
49 | print "num:",numerator, "log num:", lognum, "| denom:", denominator, "log denom:", logden, "|| log odds:", logodds
50 |
51 | # If there is a change point, then logodds will be greater than 0
52 | if logodds < 0 : return None
53 | return ( weights.argmax(), logodds )
54 |
55 | def findGaussianChangePoint( data, gammatable ):
56 | N = len( data )
57 | if N<6 : return None # can't find a cp in data this small
58 |
59 | # the denominator. This is the easy part.
60 | denom = (pi**1.5) * mpf(( N*data.var() ))**( -N/2.0 + 0.5 ) * gammatable[N]
61 |
62 | # BEGIN weight calculation
63 | # the numerator. A little trickier.
64 | weights=[0,0,0] # the change cannot have occurred in the last 3 points
65 | data2=data**2
66 |
67 | #initialize
68 | dataA=data[0:3] ; dataA2=data2[0:3] ; NA = len(dataA)
69 | dataB=data[3:] ; dataB2=data2[3:] ; NB = len(dataB)
70 | sumA=dataA.sum() ; sumsqA=dataA2.sum()
71 | sumB=dataB.sum() ; sumsqB=dataB2.sum()
72 |
73 | # first data point--this could be done in the loop but it's okay here
74 | meanA=sumA/NA ; meansumsqA = sumsqA/NA ; meanA2 = meanA**2 ; sA2=meansumsqA-meanA2
75 | meanB=sumB/NB ; meansumsqB = sumsqB/NB ; meanB2 = meanB**2 ; sB2=meansumsqB-meanB2
76 |
77 | wnumf1 = mpf(NA)**(-0.5*NA + 0.5 ) * mpf(sA2)**(-0.5*NA + 1) * gammatable[NA]
78 | wnumf2 = mpf(NB)**(-0.5*NB + 0.5 ) * mpf(sB2)**(-0.5*NB + 1) * gammatable[NB]
79 | wdenom = (sA2 + sB2) * (meanA2*meanB2)
80 | weights.append( (wnumf1*wnumf2)/wdenom )
81 |
82 | for i in range( 3, N-3 ):
83 | NA += 1 ; NB -= 1
84 | next = data[i]
85 | sumA += next ; sumB -= next
86 | nextsq = data2[i]
87 | sumsqA += nextsq; sumsqB -= nextsq
88 | meanA=sumA/NA ; meansumsqA = sumsqA/NA ; meanA2 = meanA**2 ; sA2=meansumsqA-meanA2
89 | meanB=sumB/NB ; meansumsqB = sumsqB/NB ; meanB2 = meanB**2 ; sB2=meansumsqB-meanB2
90 | wnumf1 = mpf(NA)**(-0.5*NA + 0.5 ) * mpf(sA2)**(-0.5*NA + 1) * gammatable[NA]
91 | wnumf2 = mpf(NB)**(-0.5*NB + 0.5 ) * mpf(sB2)**(-0.5*NB + 1) * gammatable[NB]
92 | wdenom = (sA2 + sB2) * (meanA2*meanB2)
93 | weights.append( (wnumf1*wnumf2)/wdenom)
94 | weights.extend( [0,0] ) # the change cannot have occurred at the last 2 points
95 | weights=array(weights)
96 | # END weight calculation
97 |
98 | num = 2.0**2.5 * abs(data.mean()) * weights.mean()
99 | logodds = log( num ) - log( denom )
100 | print "num:", num, "log num:", log(num), "| denom:", denom, "log denom:", log(denom), "|| log odds:", logodds
101 |
102 | # If there is a change point, then logodds will be greater than 0
103 | if logodds < 0 : return None
104 | return ( weights.argmax(), logodds )
105 |
106 | class ChangePointDetector:
107 | def __init__( self, data, function, table=None ):
108 | self.data = data
109 | self.datalen = len( self.data )
110 | self.function = function
111 | self.table=table
112 | self.changepoints = []
113 | self.logodds = {}
114 | self.niter = 0
115 | self.maxiter = 1000000 # just in case
116 |
117 | def nchangepoints( self ):
118 | return len( self.changepoints )
119 |
120 | def split_init( self, verbose=False ):
121 | self.split( 0, self.datalen, verbose )
122 |
123 | def split( self, start, end, verbose=False ):
124 | if self.niter > self.maxiter :
125 | print "Change point detection error: number of iterations exceeded"
126 | print "If this is the right result, you may need to increase"
127 | print "ChangePointDetector.maxiter (currently %d)" % self.maxiter
128 | return
129 | self.niter += 1
130 | if verbose:
131 | print "\nIteration %d" % self.niter
132 | print "Trying to split the segment:", self.data[start:end], "(data from %d to %d)" % ( start, end)
133 | print self.data[start:end]
134 |
135 | # try to find a change point in the data segment
136 | try:
137 | result = self.function( self.data[ start: end ], self.table )
138 | except TypeError:
139 | print "trying to test data from %d to %d failed" % ( start,end )
140 | #print self.data
141 | raise
142 |
143 | # otherwise, store the cp and call self.split on the two ends
144 | if result is not None :
145 | try: # fails if only one value is returned
146 | logodds = result[1]
147 | self.logodds[ start+result[0] ] = logodds
148 | result = start+result[0]
149 | except TypeError: # must mean it's one number?
150 | result += start
151 |
152 | if verbose: print "!! change point detected at %d !!" % result
153 | self.changepoints.append( result )
154 | self.split( start, result, verbose )
155 | self.split( result+1, end, verbose )
156 |
157 | def sort( self ): self.changepoints.sort()
158 |
159 | # display the change points
160 | def show( self ): print self.changepoints
161 |
162 | # show the change points along with the log odds
163 | def showall( self ):
164 | for i in range( len( self.changepoints ) ) :
165 | changepoint = self.changepoints[i]
166 | try:
167 | logodds = self.logodds[ changepoint ]
168 | except KeyError:
169 | logodds = None
170 | print "%d (%f)" % ( changepoint, logodds )
171 |
172 | def largest_logodds( self ):
173 | return array( self.logodds.values() ).max()
174 |
--------------------------------------------------------------------------------
/cpdetect/cp_detector.py:
--------------------------------------------------------------------------------
1 | """
2 | Bayesian change point detection. Implementation of Ensign And Pande, J. Phys. Chem. B 2010, 114, 280-292 for
3 | a normal and log-normal distribution
4 |
5 | Author: Chaya D. Stern
6 | """
7 |
8 | import numpy as np
9 | import copy
10 | from cpdetect.utils import logger
11 | import time
12 | import pandas as pd
13 | import math
14 | from scipy.special import gammaln
15 | from collections import OrderedDict
16 | from scipy.misc import logsumexp
17 |
18 |
19 | class Detector(object):
20 | """
21 | Bayesian change point detection.
22 |
23 | Attributes
24 | ----------
25 | change_points: dict of pandas dataframes
26 | This dictionary maps the trajectory to time point split, the log_odds, start, end and probability of the time points
27 | no_split: dict of pandas dataframes
28 | This dictionary maps trajectory to time points that have a peak in probability of splitting but is below the threshold.
29 | This is used for the refinement step
30 | state_emission: dict of pandas dataframes
31 | This dict maps trajectories to the sampel mu and sigma of the split segment. It's used to generate the step function
32 | log_gamme: log gamma for all Ns
33 | threshold: int
34 | log odds threshold to split. Default is 0. The lower the threshold, the more sensitive the splitting
35 | step_function: dict of arrays
36 | maps trajectory to numpy array of step function
37 | refined_change_point: dict of pandas dataframes
38 | If the refinment function is used, this dictionary will be populated with the new splits found
39 | window_size: int
40 | How many datapoints to include in the moving window. Defualt is None. When the default is none, no moving window
41 | is used
42 | stride: int
43 | The stride of the moving window. If None, no moving window is used. Default is None
44 |
45 | """
46 |
47 | def __init__(self, observations, distribution, log_odds_threshold=0, window_size=None, stride=None):
48 | """
49 |
50 | :param observations: list of numpy arrays
51 | list of observation trajectories
52 | :param distribution: str
53 | distribution of process (log_normal or normal)
54 | """
55 | self._observations = copy.deepcopy(observations)
56 | self._nobs = len(observations)
57 | self._Ts = [len(o) for o in observations]
58 | self.change_points = {} # Dictionary containing change point time and its likelihood
59 | self.no_split = {}
60 | self.state_emission = {} # Dictionary containing state's mean and sigma for segment
61 | self.loggamma = [-99, -99, -99]
62 | self.threshold = log_odds_threshold
63 | self.step_function = {}
64 | self.refined_change_points = {} # Dictionary containing new change points found during refinement.
65 |
66 | self.window_size = window_size
67 | self.stride = stride
68 | self.moving_window = False
69 | if window_size is not None:
70 | self.moving_window = True
71 |
72 | if distribution == 'log_normal':
73 | self._distribution = LogNormal()
74 | self.distribution = 'log_normal'
75 | elif distribution == 'normal' or distribution == 'gaussian':
76 | self._distribution = Normal()
77 | self.distribution = 'normal'
78 | else:
79 | raise ValueError('Use log_normal or normal distribution. I got something else')
80 |
81 | # Generate gamma table
82 | self._generate_loggamma_table()
83 |
84 | @property
85 | def nobservations(self):
86 | """ Number of observation trajectories """
87 | return self._nobs
88 |
89 | @property
90 | def observation_lengths(self):
91 | """ Return lengths of trajectories"""
92 | return self._Ts
93 |
94 | def _normal_lognormal_bf(self, obs):
95 | """
96 | Calculate Bayes factor P(D|H_2) / P(D|H_1) for normal or log-normal data
97 |
98 | :parameter:
99 | obs: np.array
100 | segment of trajectory to calculate Bayes factor
101 | :return:
102 | ts: int
103 | time point for split (argmax)
104 | log_odds: float
105 |
106 | """
107 | n = len(obs)
108 | if n < 6:
109 | logger().debug('Segment is less than 6 points')
110 | return None # can't find a cp in data this small
111 |
112 | # Calculate mean and var
113 | mean, var = self._distribution.mean_var(obs)
114 |
115 | # the denominator. This is the easy part.
116 | denom = 1.5*np.log(np.pi) + (-n/2.0 + 0.5)*(np.log(n*var)) + self.loggamma[n]
117 |
118 | # BEGIN weight calculation
119 | # the numerator. A little trickier.
120 | weights = [0, 0, 0] # the change cannot have occurred in the last 3 points
121 |
122 | for i in range(3, n-2):
123 | data_a = obs[0:i]
124 | n_a = len(data_a)
125 | data_b = obs[i:]
126 | n_b = len(data_b)
127 |
128 | mean_a, var_a = self._distribution.mean_var(data_a)
129 | mean_b, var_b = self._distribution.mean_var(data_b)
130 |
131 | mean_a2 = mean_a**2
132 | mean_b2 = mean_b**2
133 |
134 | wnumf1 = (-0.5*n_a + 0.5)*np.log(n_a) + (-0.5*n_a + 1)*np.log(var_a) + self.loggamma[n_a]
135 | wnumf2 = (-0.5*n_b + 0.5)*np.log(n_b) + (-0.5*n_b + 1)*np.log(var_b) + self.loggamma[n_b]
136 |
137 | wdenom = np.log(var_a + var_b) + np.log(mean_a2*mean_b2)
138 |
139 | weights.append((wnumf1 + wnumf2) - wdenom)
140 |
141 | weights.extend([0, 0]) # the change cannot have occurred at the last 2 points
142 | weights = np.array(weights)
143 | # END weight calculation
144 | num = 2.5*np.log(2.0) + np.log(abs(mean)) + weights.mean()
145 | log_odds = num - denom
146 | # Replace points where change cannot occur with negative infinity so that they cannot be argmax
147 | weights[0] = weights[1] = weights[2] = weights[-1] = weights[-2] = -np.inf
148 | logger().debug(' log num: ' + str(num))
149 | logger().debug(' denom: ' + str(denom))
150 | logger().debug(' log odds: ' + str(log_odds))
151 |
152 | norm = logsumexp(weights[3:-2])
153 | normalized_prob = np.exp(weights[3:-3] - norm)
154 |
155 | # If there is a change point, then logodds will be greater than 0
156 | # Check for nan. This comes up if using log normal for a normal distribution.
157 | if math.isnan(log_odds):
158 | raise ValueError('Are you using the correct distribution?')
159 | if log_odds < self.threshold:
160 |
161 | logger().debug(' Log Odds: ' + str(log_odds) + ' is less than threshold ' + str(self.threshold) +
162 | '. No change point found')
163 | return normalized_prob, weights.argmax(), log_odds, None
164 |
165 | return normalized_prob, weights.argmax(), log_odds
166 |
167 | def _generate_loggamma_table(self):
168 | """
169 | calculate log gamma for all N
170 | """
171 | for i in range(3, max(self._Ts) + 1):
172 | self.loggamma.append(gammaln(0.5*i - 1))
173 |
174 | def detect_cp(self):
175 | """
176 | Bayesian detection of Intensity changes. This function detects the changes, their timepoints and then
177 | finds the state emission for each segment to draw the step function
178 | """
179 |
180 | logger().info('=======================================')
181 | logger().info('Running change point detector')
182 | logger().info('=======================================')
183 | logger().info(' input observations: '+str(self.nobservations)+ ' of length ' + str(self.observation_lengths))
184 |
185 | initial_time = time.time()
186 |
187 | for k in range(self._nobs):
188 | logger().info('Running cp detector on traj ' + str(k))
189 | logger().info('---------------------------------')
190 | self.change_points['traj_%s' %str(k)] = pd.DataFrame(columns=['ts', 'log_odds', 'start_end'])
191 | self.change_points['traj_%s' % str(k)]['ts'] = self.change_points['traj_%s' %str(k)]['ts'].astype(int)
192 | self.no_split['traj_%s' %str(k)] = pd.DataFrame(columns=['ts', 'log_odds', 'start_end'])
193 | self.no_split['traj_%s' % str(k)]['ts'] = self.no_split['traj_%s' %str(k)]['ts'].astype(int)
194 | obs_full = self._observations[k]
195 | if self.window_size:
196 | # Iterate splitting algorithm over window
197 | chunks = (obs_full.shape[0]-self.window_size)/self.stride + 1
198 | indexer = np.arange(self.window_size)[None, :] + self.stride*np.arange(int(chunks))[:, None]
199 | observations = obs_full[indexer]
200 | for obs, index in zip(observations, indexer):
201 | self._split(obs, 0, self.window_size-1, k, indexer=index)
202 | else:
203 | self._split(obs_full, 0, self.observation_lengths[k], k)
204 | logger().info('Generating step fucntion')
205 | logger().info('---------------------------------')
206 | self._generate_step_function(obs_full, k)
207 |
208 | final_time = time.time()
209 |
210 | logger().info('Elapsed time: ' + str(final_time-initial_time))
211 |
212 | def _split(self, obs, start, end, itraj, indexer=None):
213 | """
214 | This function takes an array of observations and checks if it should be split
215 |
216 | :param obs: np.array
217 | trajectory to check for change point
218 | :param start: int
219 | start of segment to check for change point
220 | :param end: int
221 | end of segment
222 | :param itraj: int
223 | index of trajectory
224 | """
225 | # recursive function to find all ts and logg odds
226 | logger().debug(' Trying to split segment start at ' + str(start) + ' end ' + str(end))
227 | result = self._normal_lognormal_bf(obs[start:end])
228 | if result is None:
229 | logger().debug(" Can't split segment with less than 6 points.")
230 | return
231 | elif result[-1] is None:
232 | ts = start + result[1]
233 | if indexer is not None:
234 | ts = indexer[ts]
235 | start = indexer[start]
236 | end = indexer[end]
237 | logger().debug(" Can't split segment start at " + str(start) + " end at " + str(end))
238 | self.no_split['traj_%s' % str(itraj)] = self.no_split['traj_%s' % str(itraj)].append({'ts': ts,
239 | 'log_odds': result[2], 'start_end': (start, end), 'prob_ts': result[0]}, ignore_index=True)
240 | return
241 | else:
242 | log_odds = result[-1]
243 | ts = start + result[1]
244 | prob_ts = result[0]
245 | if indexer is None:
246 | self.change_points['traj_%s' % str(itraj)] = self.change_points['traj_%s' % str(itraj)].append(
247 | {'ts': ts, 'log_odds': log_odds, 'start_end': (start, end), 'prob_ts': prob_ts}, ignore_index=True)
248 | else:
249 | self.change_points['traj_%s' % str(itraj)] = self.change_points['traj_%s' % str(itraj)].append(
250 | {'ts': indexer[ts], 'log_odds': log_odds, 'start_end': (indexer[start], indexer[end]), 'prob_ts': prob_ts}, ignore_index=True)
251 |
252 | logger().info(' Found a new change point at: ' + str(ts) + '!!')
253 | self._split(obs, start, ts, itraj, indexer)
254 | self._split(obs, ts, end, itraj, indexer)
255 |
256 | def _generate_step_function(self, obs, itraj, refined=False):
257 | """Draw step function based on sample mean
258 |
259 | :parameter obs
260 | trajectory
261 | :parameter itraj: int
262 | index of trajectory
263 | """
264 |
265 | self.state_emission['traj_%s' % str(itraj)] = pd.DataFrame(columns=['partition', 'sample_mu', 'sample_sigma'])
266 |
267 | # First sort ts of traj
268 | ts = self.change_points['traj_%s' % str(itraj)]['ts']
269 | if refined:
270 | ts = ts.append(self.refined_change_points['traj_{}'.format(itraj)]['ts'])
271 | if len(ts) == 0:
272 | logger().info('No change point was found')
273 | self.step_function['traj_%s' % str(itraj)] = np.ones(self.observation_lengths[itraj]-1)
274 | mean, var = self._distribution.mean_var(obs)
275 | self.step_function['traj_%s' % str(itraj)] = self.step_function['traj_%s' % str(itraj)]*np.exp(mean)
276 | return
277 | ts = ts.drop_duplicates().sort_values().values
278 | # populate data frame with partitions, sample mean and sigma
279 | partitions = [(0, int(ts[0]))]
280 | mean, var = self._distribution.mean_var(obs[0:ts[0]])
281 | means = [mean]
282 | sigmas = [var]
283 | for i, j in enumerate(ts):
284 | try:
285 | partitions.append((int(j+1), int(ts[i+1])))
286 | mean, var = self._distribution.mean_var(obs[j+1:ts[i+1]])
287 | means.append(mean)
288 | sigmas.append(var)
289 |
290 | except IndexError:
291 | partitions.append((int(ts[-1]+1), int(self.observation_lengths[itraj]-1)))
292 | mean, var = self._distribution.mean_var(obs[ts[-1]+1:len(obs)-1])
293 | means.append(mean)
294 | sigmas.append(var)
295 |
296 | except ValueError:
297 | pass
298 | self.state_emission['traj_%s' % str(itraj)]['partition'] = partitions
299 | self.state_emission['traj_%s' % str(itraj)]['sample_mu'] = means
300 | self.state_emission['traj_%s' % str(itraj)]['sample_sigma'] = sigmas
301 |
302 | # generate step function
303 | self.step_function['traj_%s' % str(itraj)] = np.ones(self.observation_lengths[itraj])
304 | for index, row in self.state_emission['traj_%s' % str(itraj)].iterrows():
305 | self.step_function['traj_%s' % str(itraj)][row['partition'][0]:row['partition'][1]+1] = \
306 | np.exp(row['sample_mu'])
307 |
308 | def refinement(self, threshold=0, split_window=50, reject_window=10):
309 | """
310 | This function goes through all rejected splits and recalculates the Bayes factor on splits of the segments. The
311 | splits are given by the ts + and - the split window. If a new change point is found, it will be accepted given
312 | that the log odds are above the threshold and there is no other predicted change point within the reject window.
313 |
314 | Parameters
315 | ----------
316 | threshold : float or int
317 | log odds threshold to accept a split
318 | split_window : int
319 | how many points out of rejected split to calculate Bayes Factor
320 | reject_window : int
321 | The window for which another predicted change point will be considered equal to a new change point
322 |
323 | """
324 |
325 | for t in range(self.nobservations):
326 | self.refined_change_points['traj_{}'.format(str(t))] = pd.DataFrame(columns=['ts', 'log_odds', 'start_end'])
327 | self.refined_change_points['traj_%s' % str(t)]['ts'] = self.refined_change_points['traj_%s'
328 | % str(t)]['ts'].astype(int)
329 | predicted_ts = np.array(self.change_points['traj_{}'.format(t)]['ts'])
330 | for index, row in self.no_split['traj_{}'.format(t)].iterrows():
331 | start, end = row['start_end']
332 | ts = row['ts']
333 | new_splits = [(start, ts + split_window), (ts - split_window, end)]
334 | obs1 = self._observations[t][new_splits[0][0]:new_splits[0][1]]
335 | obs2 = self._observations[t][new_splits[1][0]:new_splits[1][1]]
336 | bf = self._normal_lognormal_bf(obs1)
337 |
338 | if bf is not None and bf[2] > threshold:
339 | new_ts = bf[1] + new_splits[0][0]
340 | if not np.any((predicted_ts < new_ts + reject_window) & (predicted_ts > new_ts - reject_window)):
341 | logger().info('Found a new change point in traj {} at {}'.format(t, new_ts))
342 | self.refined_change_points['traj_{}'.format(str(t))] = self.refined_change_points[
343 | 'traj_{}'.format(t)].append({'ts': new_ts, 'log_odds': bf[2], 'start_end': new_splits[0]},
344 | ignore_index=True)
345 |
346 | bf = self._normal_lognormal_bf(obs2)
347 | if bf is not None and bf[2] > threshold:
348 | new_ts = bf[1] + new_splits[1][0]
349 | if not np.any((predicted_ts < new_ts + reject_window) & (predicted_ts > new_ts - reject_window)):
350 | logger().info('Found a new change point in traj {} at {}'.format(t, new_ts))
351 | self.refined_change_points['traj_{}'.format(str(t))] = self.refined_change_points[
352 | 'traj_{}'.format(t)].append({'ts': new_ts, 'log_odds': bf[2], 'start_end': new_splits[1]},
353 | ignore_index=True)
354 |
355 | self.refined_change_points['traj_{}'.format(str(t))]['ts'].drop_duplicates(inplace=True)
356 |
357 | def regenerate_step_function(self):
358 | for t in range(self.nobservations):
359 | obs = self._observations[t]
360 | self._generate_step_function(obs, t, refined=True)
361 |
362 | def to_csv(self, filename=None):
363 | """
364 | export change_points data frame to csv file
365 | :parameter:
366 | filename: str
367 |
368 | :return:
369 | csv if no filename given. Otherwise, saves csv file
370 | """
371 | frames = []
372 | keys = []
373 | for i in self.change_points:
374 | keys.append(i)
375 | frames.append(self.change_points[i])
376 | all_f = pd.concat(frames, keys=keys)
377 |
378 | if filename:
379 | all_f.to_csv(filename)
380 | else:
381 | return all.to_csv()
382 |
383 |
384 | class LogNormal(object):
385 |
386 | @classmethod
387 | def mean_var(cls, data):
388 | """
389 | calculate log normal mean and variance (loc and scale)
390 | :parameter:
391 | data: np.array
392 | data points to calculate mean and var
393 |
394 | :return: (float, float)
395 | loc, scale of data
396 | """
397 | n = len(data)
398 | logx = np.log(data)
399 | loc = logx.sum()/n
400 | scale = ((logx - loc)**2).sum()/n
401 | return loc, scale
402 |
403 |
404 | class Normal(object):
405 |
406 | @classmethod
407 | def mean_var(cls, data):
408 | return data.mean(), data.var()
409 |
--------------------------------------------------------------------------------
/attic/trajectory.dat:
--------------------------------------------------------------------------------
1 |
2 | (lp1
3 | F50.275547440231023
4 | aF51.791625880721611
5 | aF49.684985565970329
6 | aF49.34984584006699
7 | aF49.98997797686696
8 | aF49.910605202345167
9 | aF49.424888610493937
10 | aF50.756859876700645
11 | aF51.434684516073261
12 | aF50.335890278414119
13 | aF50.112353522081207
14 | aF49.07397590749779
15 | aF52.977858133439327
16 | aF48.131329833779795
17 | aF50.01719929909904
18 | aF49.294792261668213
19 | aF49.927953396299571
20 | aF49.77909330257993
21 | aF48.687443500916302
22 | aF50.413377091913922
23 | aF50.408202102849842
24 | aF49.587981442294733
25 | aF51.760343541992441
26 | aF48.97774976067813
27 | aF50.03565460113002
28 | aF51.384082986960934
29 | aF50.487410581613794
30 | aF53.797596218048014
31 | aF51.273423517441444
32 | aF49.713542574044467
33 | aF51.024844612103763
34 | aF49.862855611535878
35 | aF51.497052375017347
36 | aF51.504675607088977
37 | aF52.285486785054623
38 | aF51.270876547734467
39 | aF51.119322487040968
40 | aF49.076429194165797
41 | aF50.923136918523603
42 | aF50.088701991472618
43 | aF50.96626998956291
44 | aF51.159564401584269
45 | aF49.081831876100438
46 | aF51.992765646470708
47 | aF50.807212240900796
48 | aF49.073334927494336
49 | aF50.155274098461973
50 | aF50.516952043294914
51 | aF51.988869889685468
52 | aF48.362197057558731
53 | aF52.209408077908776
54 | aF49.547030571852559
55 | aF52.652555954924601
56 | aF51.535502602697981
57 | aF48.788054454419559
58 | aF51.99556290033685
59 | aF52.476753906877072
60 | aF50.385929915634911
61 | aF49.132989179901564
62 | aF49.844442753656963
63 | aF50.176363436294061
64 | aF49.735811707754664
65 | aF52.081509293525961
66 | aF50.162430834554797
67 | aF49.103887525704593
68 | aF49.687860832844279
69 | aF50.879559114084906
70 | aF49.2296464422556
71 | aF52.154099831027324
72 | aF51.87678175945608
73 | aF41.551577051489268
74 | aF41.906293807405682
75 | aF42.290129644194607
76 | aF42.270266298581639
77 | aF41.969858045002852
78 | aF41.763508499310454
79 | aF43.2424518726245
80 | aF42.209490489110422
81 | aF40.930190373551966
82 | aF42.603447065931945
83 | aF42.499957484502566
84 | aF42.924243503938094
85 | aF43.267414572936964
86 | aF43.603526396346226
87 | aF42.308979417071257
88 | aF41.860177674481456
89 | aF43.300368781447702
90 | aF42.015025004961352
91 | aF43.322751068355416
92 | aF43.110788448704767
93 | aF43.296587081841643
94 | aF41.277445900769585
95 | aF42.008540368506715
96 | aF42.616072250154545
97 | aF42.994395355450564
98 | aF35.985407968035112
99 | aF39.040051961044838
100 | aF39.472907236087352
101 | aF37.067217280984678
102 | aF37.737882096051841
103 | aF37.289538753841306
104 | aF37.375182752998789
105 | aF38.202387055291339
106 | aF38.420615626723198
107 | aF38.742855666959798
108 | aF37.903433588364521
109 | aF37.171711338048453
110 | aF36.536314600090662
111 | aF38.515521693474277
112 | aF36.79792623040597
113 | aF36.980741632652446
114 | aF37.208590759545814
115 | aF37.727559064579978
116 | aF37.204060900189809
117 | aF36.007831995727251
118 | aF37.488974071786537
119 | aF37.563812838978173
120 | aF37.632370971729415
121 | aF38.263420418948044
122 | aF36.380794151432383
123 | aF38.583654278910458
124 | aF38.48135286680612
125 | aF36.134033105060283
126 | aF38.270791828660521
127 | aF37.993546185413543
128 | aF38.859066656751125
129 | aF38.128547358591675
130 | aF36.175614301283638
131 | aF40.123345427416574
132 | aF36.211935444320915
133 | aF38.531909423597902
134 | aF38.382873091313087
135 | aF37.207267242245337
136 | aF38.512877921713581
137 | aF38.753347236949082
138 | aF36.635808051350743
139 | aF37.445421281879668
140 | aF36.852113343637591
141 | aF38.524816726295398
142 | aF39.253073810872955
143 | aF38.862504486261301
144 | aF37.754343074776834
145 | aF36.481745059537211
146 | aF36.912168570986388
147 | aF28.171895300846693
148 | aF32.808391998170599
149 | aF28.623126263223519
150 | aF32.038652515152563
151 | aF34.060517509182581
152 | aF32.566085509360242
153 | aF34.355177762300634
154 | aF34.188905817435504
155 | aF33.101329112071625
156 | aF35.84608667097659
157 | aF32.096127896366127
158 | aF31.793424632456983
159 | aF31.111993820671007
160 | aF35.545146673462831
161 | aF29.890342413695798
162 | aF31.571600297155296
163 | aF34.346921071066184
164 | aF31.784367089630823
165 | aF29.036211829009055
166 | aF32.631933261005777
167 | aF29.465090453521782
168 | aF30.718691670264388
169 | aF33.863517211722147
170 | aF32.474088794592667
171 | aF33.137513297479025
172 | aF31.269995636212535
173 | aF33.153956574137851
174 | aF33.052047475498647
175 | aF33.43155834353874
176 | aF30.084429259578439
177 | aF31.930457045138247
178 | aF32.835609106644178
179 | aF28.921836588151244
180 | aF31.961267966524296
181 | aF30.624090324313581
182 | aF32.373055715322309
183 | aF30.252360288684276
184 | aF29.857617527785777
185 | aF33.056214228522407
186 | aF33.589789539485018
187 | aF30.870865697841396
188 | aF30.090379278937728
189 | aF34.998072570027453
190 | aF33.242880772547849
191 | aF32.710080067542627
192 | aF31.865897190004343
193 | aF33.520367804793246
194 | aF33.673934522399428
195 | aF34.683235382453738
196 | aF33.202561513106794
197 | aF30.792932052618553
198 | aF34.367426496358171
199 | aF35.73683278714109
200 | aF30.897156533998704
201 | aF31.736591206237939
202 | aF33.059106854951203
203 | aF32.780607764334903
204 | aF36.22484467791336
205 | aF30.875626523257175
206 | aF29.788664120753928
207 | aF35.226370097252428
208 | aF32.495123361535732
209 | aF30.279450650883579
210 | aF33.539927908009844
211 | aF31.805856870129453
212 | aF32.657247333454386
213 | aF34.168472579455589
214 | aF31.045704625327133
215 | aF33.655983108392732
216 | aF30.461443010295259
217 | aF34.586718773261602
218 | aF31.26066315010365
219 | aF32.505319249266925
220 | aF31.403073900677867
221 | aF33.623482291446848
222 | aF32.567793088342356
223 | aF32.466725300279371
224 | aF30.812453426490368
225 | aF34.673059874428574
226 | aF33.566193146752191
227 | aF33.369242964694102
228 | aF32.507718074488828
229 | aF33.308935811763838
230 | aF30.603800041682454
231 | aF33.898245658167482
232 | aF32.976790146911434
233 | aF32.807256051983593
234 | aF33.425996405695422
235 | aF32.251349417424819
236 | aF31.258440222397862
237 | aF30.807300842007493
238 | aF31.762565223934391
239 | aF32.672236370583725
240 | aF31.569850642624044
241 | aF31.817077108972736
242 | aF30.839644941068904
243 | aF31.692472750258034
244 | aF29.131848489151398
245 | aF29.709112287610154
246 | aF30.602010136483095
247 | aF33.801398077126585
248 | aF32.51561246087801
249 | aF33.624590786535727
250 | aF33.188448905913866
251 | aF32.013987797176448
252 | aF32.521937657711959
253 | aF34.162448384074665
254 | aF32.236086192184189
255 | aF32.982600581811376
256 | aF32.223388466183046
257 | aF33.909777034278846
258 | aF30.932317175503464
259 | aF35.42460902139247
260 | aF32.882587077584844
261 | aF36.424614037330542
262 | aF28.06356006918109
263 | aF31.525851609703778
264 | aF31.766983422122983
265 | aF32.761476718208073
266 | aF28.294515311005494
267 | aF31.567463795909045
268 | aF34.040650689486149
269 | aF33.182583364394873
270 | aF32.492762547596826
271 | aF30.427663511894664
272 | aF31.132882594746103
273 | aF34.705112370686415
274 | aF32.590085239776663
275 | aF30.001250692672549
276 | aF32.358081296528205
277 | aF31.791903595580646
278 | aF32.969805859772812
279 | aF31.792279199628446
280 | aF31.179819756612211
281 | aF31.559360918015873
282 | aF32.921798661417085
283 | aF31.191324996378722
284 | aF31.872213185654282
285 | aF32.921789822359962
286 | aF32.082359655644176
287 | aF32.006921719201415
288 | aF30.795524065860238
289 | aF33.950482593307882
290 | aF36.614809401206287
291 | aF32.621167293204394
292 | aF32.652981437921085
293 | aF31.400716498967384
294 | aF30.608379755848652
295 | aF31.733740501147928
296 | aF33.406747266728452
297 | aF32.522230750316268
298 | aF31.904934965616956
299 | aF33.608987042172082
300 | aF32.098840929841145
301 | aF35.160805151381339
302 | aF28.237031051218988
303 | aF30.437018790162256
304 | aF30.616621601979922
305 | aF31.671145411003298
306 | aF32.00564690612044
307 | aF32.406966067260498
308 | aF31.196479978893105
309 | aF31.182441176105481
310 | aF31.142256773351452
311 | aF32.011036087690748
312 | aF31.035681646342063
313 | aF32.137925943982339
314 | aF32.169530325936698
315 | aF33.821064147834925
316 | aF31.653970582869157
317 | aF30.762069193778288
318 | aF31.759738700885197
319 | aF29.459485299742191
320 | aF31.723721541840867
321 | aF30.225509461070239
322 | aF33.576124243535382
323 | aF34.628962053546203
324 | aF31.878635752106309
325 | aF34.999533655987243
326 | aF33.198717723540028
327 | aF31.962779152937127
328 | aF33.479125713818284
329 | aF31.252400109842331
330 | aF34.16880221861242
331 | aF32.356120428603866
332 | aF35.477428565959052
333 | aF32.666974716337769
334 | aF34.514158334605597
335 | aF32.499177654706997
336 | aF32.382214140029923
337 | aF32.007346430033095
338 | aF32.514855855563695
339 | aF35.351418363895228
340 | aF30.38125442469136
341 | aF31.547426103706155
342 | aF31.763840287520818
343 | aF34.408971642168787
344 | aF31.711670827723427
345 | aF31.165060801307106
346 | aF31.861593595917711
347 | aF33.3642391236178
348 | aF32.876427116588076
349 | aF31.574949383727308
350 | aF33.224766270309075
351 | aF31.694223599655707
352 | aF28.979333579909419
353 | aF28.950855960541357
354 | aF31.477813906206546
355 | aF31.354893696928141
356 | aF31.468502913904842
357 | aF31.415188081150042
358 | aF33.560926336131054
359 | aF29.578231248285206
360 | aF32.648215035977287
361 | aF30.072797267572028
362 | aF32.562108159917663
363 | aF30.888087188823913
364 | aF29.63411475678102
365 | aF31.23104305695691
366 | aF31.695922999356497
367 | aF32.866170187899456
368 | aF33.971590457370077
369 | aF30.232724431466789
370 | aF32.472953638244491
371 | aF31.345942485013179
372 | aF33.60973011787123
373 | aF30.839444648285994
374 | aF32.009073632919318
375 | aF34.514453998756736
376 | aF30.285560451894447
377 | aF33.179148586733234
378 | aF29.43321574380402
379 | aF34.222160440627952
380 | aF31.053677366123118
381 | aF30.662630245725392
382 | aF31.739063479551774
383 | aF31.28893127756195
384 | aF30.625507652922739
385 | aF31.716336654522586
386 | aF32.717181253339753
387 | aF31.441862687242921
388 | aF32.06271541710742
389 | aF29.333074742777541
390 | aF28.31266502241365
391 | aF30.726286065236824
392 | aF31.847861112515815
393 | aF32.930504401403667
394 | aF32.303738540617211
395 | aF34.006003261568587
396 | aF29.387602602406069
397 | aF29.753052141616614
398 | aF33.259757579741951
399 | aF32.954945812948154
400 | aF32.758292268866981
401 | aF31.669700710107787
402 | aF32.316034163107609
403 | aF31.662617799577212
404 | aF33.457101581870965
405 | aF32.689842616992941
406 | aF32.881719390423392
407 | aF31.335123210605346
408 | aF35.198123806784523
409 | aF31.811593227836109
410 | aF30.851094900335976
411 | aF31.325900373148695
412 | aF31.531721002554001
413 | aF30.508731136600158
414 | aF33.179427485360677
415 | aF32.511264527663712
416 | aF34.17582462364102
417 | aF32.189268429025788
418 | aF31.623634382396659
419 | aF31.049782737940937
420 | aF31.726618566130973
421 | aF35.922891883468822
422 | aF36.031371258933447
423 | aF37.603885319972967
424 | aF34.500375557605018
425 | aF37.249904194968849
426 | aF36.644675074854163
427 | aF34.91553184325312
428 | aF33.985264996108548
429 | aF39.04434373988137
430 | aF35.633721978186983
431 | aF37.336338453438586
432 | aF35.336668118565207
433 | aF37.804280430773673
434 | aF37.5311304110378
435 | aF37.629505608338306
436 | aF38.202554752080594
437 | aF36.725132378645377
438 | aF36.607610590303878
439 | aF35.428069646607192
440 | aF35.564331848037057
441 | aF37.653893119471064
442 | aF37.075390642162738
443 | aF37.487568938287559
444 | aF37.46333280974558
445 | aF36.895259359543473
446 | aF36.452730596123104
447 | aF38.179814915389798
448 | aF35.585417670183965
449 | aF37.462348789259963
450 | aF38.041533417707313
451 | aF37.672808005366363
452 | aF37.757576226792388
453 | aF36.762712537751398
454 | aF36.71485478876015
455 | aF36.396703154757127
456 | aF38.164273698603161
457 | aF36.536895842660684
458 | aF36.813334843403027
459 | aF35.128691343844352
460 | aF36.600761731739297
461 | aF35.836758290321974
462 | aF45.792704357023581
463 | aF44.081430922890341
464 | aF44.932862333185
465 | aF44.758151273324692
466 | aF44.711076336030196
467 | aF43.342565914952637
468 | aF42.653293725000815
469 | aF44.081812748217907
470 | aF44.246991183794016
471 | aF44.833953221643185
472 | aF44.814593015181096
473 | aF44.921088954092518
474 | aF43.105882913168273
475 | aF44.07788875503924
476 | aF45.773594829942233
477 | aF45.123988385980468
478 | aF42.81102917685638
479 | aF45.083250962848055
480 | aF45.889935501343054
481 | aF43.970893949956533
482 | aF43.605446952654368
483 | aF44.659163365135335
484 | aF43.616386070129337
485 | aF45.169572443451401
486 | aF42.411604787467319
487 | aF45.653383054665724
488 | aF45.727812070961086
489 | aF48.090960649811784
490 | aF43.258652397049325
491 | aF46.574913862028438
492 | aF44.064728219651499
493 | aF44.526617035224319
494 | aF44.673318402550528
495 | aF45.82399102126702
496 | aF41.204922913243507
497 | aF41.697929908760173
498 | aF45.832676751333636
499 | aF44.39492289865936
500 | aF48.989115791685535
501 | aF47.923089330906187
502 | aF47.011697065216715
503 | aF46.851577803994886
504 | aF48.71175839537743
505 | aF47.378384958476055
506 | aF46.515883235018798
507 | aF46.666127554429963
508 | aF48.279416994357987
509 | aF47.261741225855424
510 | aF45.726206251812563
511 | aF46.343946131940555
512 | aF46.685545132135367
513 | aF47.826134258634646
514 | aF46.577097438863497
515 | aF46.607561496186236
516 | aF44.436964249124422
517 | aF45.98615327229583
518 | aF48.162199279446945
519 | aF48.099768637981512
520 | aF46.404347582055202
521 | aF47.062582596517444
522 | aF49.010425140906634
523 | aF48.060858515695465
524 | aF45.838169389484683
525 | aF45.026745848746927
526 | aF48.82181263746115
527 | aF47.292950804364047
528 | aF45.02843671584867
529 | aF48.015632329107888
530 | aF49.718630460894381
531 | aF47.875859423306295
532 | aF46.995875995011637
533 | aF46.476718637847341
534 | aF46.415929887636615
535 | aF46.725490209844544
536 | aF46.770826277197543
537 | aF45.768486081615414
538 | aF47.040590610664836
539 | aF44.646243775708307
540 | aF46.963633027685169
541 | aF46.6847858738208
542 | aF46.7199458845026
543 | aF48.341804966373651
544 | aF47.982780113595958
545 | aF47.520961745753631
546 | aF48.38153051254622
547 | aF48.87876468368372
548 | aF46.242692271515942
549 | aF47.545837925912252
550 | aF47.118590716447592
551 | aF47.510442832285563
552 | aF47.901200935888248
553 | aF45.923833444489006
554 | aF47.604515096344635
555 | aF46.292285263914188
556 | aF46.81690585034881
557 | aF46.821380348071216
558 | aF47.697255595105425
559 | aF46.32075714055982
560 | aF49.390815470436323
561 | aF47.293443648765368
562 | aF46.436200035320027
563 | aF46.452132919062016
564 | aF45.998757114015007
565 | aF46.287687319878842
566 | aF46.259995933450803
567 | aF46.576464453909729
568 | aF48.011206685045352
569 | aF47.438032133597808
570 | aF47.375052385008686
571 | aF46.917893843373449
572 | aF47.637451839989644
573 | aF47.582453383075652
574 | aF47.658665963942859
575 | aF47.100677802844245
576 | aF46.16145772759382
577 | aF48.020766535291742
578 | aF46.394606029825361
579 | aF47.87898517678525
580 | aF46.55659262446968
581 | aF45.794714981326813
582 | aF46.464720913563795
583 | aF46.685432373182415
584 | aF46.406962422067551
585 | aF47.24065219156499
586 | aF47.596560722662495
587 | aF47.985709951512753
588 | aF47.310054970737255
589 | aF48.468886072244658
590 | aF48.329020187665414
591 | aF48.588142375065118
592 | aF46.993147941591779
593 | aF47.275572792354737
594 | aF48.40186423037067
595 | aF46.917385333087921
596 | aF48.239011399612927
597 | aF48.357541559857353
598 | aF46.739231725487201
599 | aF45.755957988415545
600 | aF47.926391454375931
601 | aF46.450500844473346
602 | aF46.238716042212161
603 | aF47.806006574616582
604 | aF45.033000984108227
605 | aF47.398831557421367
606 | aF46.448687978729517
607 | aF45.848005589766053
608 | aF45.801366427626149
609 | aF47.44589314008126
610 | aF49.337110298579098
611 | aF48.385019229930748
612 | aF46.619572740906548
613 | aF47.865989689470602
614 | aF46.591407864641262
615 | aF47.260099161487268
616 | aF48.611742604173969
617 | aF47.698594686509452
618 | aF46.85488005404153
619 | aF47.075893379312447
620 | aF48.342529585387759
621 | aF49.272080462014145
622 | aF48.476280427104072
623 | aF47.952306005979871
624 | aF44.987468581338128
625 | aF45.432094487375267
626 | aF47.408554655729731
627 | aF45.611740030580492
628 | aF48.372862091702281
629 | aF46.450022295649411
630 | aF48.506721316844057
631 | aF46.54269917816741
632 | aF46.645440751624534
633 | aF47.431936569752828
634 | aF46.936265616771465
635 | aF46.312455595532697
636 | aF47.171851615396598
637 | aF48.294745821885648
638 | aF47.878515306603283
639 | aF45.751258619230242
640 | aF47.282302573877189
641 | aF46.749983819301029
642 | aF46.455872264822418
643 | aF47.826276767295951
644 | aF48.501043991896097
645 | aF48.330208807673102
646 | aF47.643586865132555
647 | aF47.740674740949018
648 | aF49.390511698789446
649 | aF49.200597822846206
650 | aF48.651452201209196
651 | aF47.6208644555837
652 | aF48.314247795548781
653 | aF49.183224208359142
654 | aF50.110662697526351
655 | aF46.593108584311125
656 | aF45.839654484638324
657 | aF46.837994458098038
658 | aF45.502133777832668
659 | aF45.891823605856608
660 | aF46.842584968067861
661 | aF45.266722715476995
662 | aF45.819029559044537
663 | aF45.185181643800085
664 | aF45.424857697780666
665 | aF46.884821518714304
666 | aF45.495874016088315
667 | aF45.349027704625655
668 | aF44.54127431647364
669 | aF45.494897324964647
670 | aF45.968782884887084
671 | aF45.636884758772197
672 | aF46.401565164087579
673 | aF45.057602704107687
674 | aF45.565244598416967
675 | aF53.286041068668197
676 | aF52.803212902990687
677 | aF51.998810007676049
678 | aF55.701499827836358
679 | aF55.162652669388741
680 | aF56.151107571647934
681 | aF64.665743159790679
682 | aF63.308910182068381
683 | aF64.770559863510186
684 | aF63.977457431595809
685 | aF63.995377906703112
686 | aF66.128359651874248
687 | aF63.600610419748399
688 | aF64.494301747180955
689 | aF63.792330376366102
690 | aF63.828611286262813
691 | aF65.428905452288518
692 | aF64.425509026593758
693 | aF65.700181209322395
694 | aF62.507269894975835
695 | aF64.636369941480865
696 | aF66.384956716460835
697 | aF65.528321677698855
698 | aF64.016490623936406
699 | aF63.892018989237194
700 | aF64.558765549607003
701 | aF64.408575856479104
702 | aF63.727875231106673
703 | aF64.516958922595308
704 | aF63.52905127668916
705 | aF62.98738595529047
706 | aF64.114101049981869
707 | aF63.740811445939286
708 | aF63.474824724263641
709 | aF63.726998035843309
710 | aF63.095738245448054
711 | aF65.730204398880772
712 | aF63.939876372553542
713 | aF63.144226228654631
714 | aF63.801485483523976
715 | aF63.546899098356363
716 | aF65.810850586533036
717 | aF63.920201849525796
718 | aF63.443012470912151
719 | aF63.436203759499229
720 | aF64.096619079674099
721 | aF64.071607680030908
722 | aF62.601963066756603
723 | aF63.47210864006432
724 | aF63.937880937421475
725 | aF62.350041228670896
726 | aF63.41164288199321
727 | aF63.852772221381748
728 | aF63.51864398442477
729 | aF63.70596793993505
730 | aF63.154204397102504
731 | aF62.775437296865185
732 | aF62.796528900905685
733 | aF63.944799371466175
734 | aF63.559850754784158
735 | aF62.461252779659631
736 | aF64.228385390485954
737 | aF63.643159875404166
738 | aF62.294568421226835
739 | aF63.548675683527591
740 | aF62.646221445371232
741 | aF62.777228471954857
742 | aF62.031250633413208
743 | aF63.139352230454691
744 | aF62.921524980414944
745 | aF63.33546206144576
746 | aF63.019864289211526
747 | aF63.431712337461754
748 | aF63.611720725625567
749 | aF64.095391577655846
750 | aF63.750684019759412
751 | aF63.164925794041281
752 | aF62.459038549313526
753 | aF63.130879346400064
754 | aF63.374582113929193
755 | aF63.520780702018314
756 | aF62.260089500553249
757 | aF62.468625895951575
758 | aF63.63359196854568
759 | aF62.109224676618943
760 | aF64.124561908794902
761 | aF63.165479308257417
762 | aF63.177703041553933
763 | aF61.986173839842635
764 | aF64.090697121572219
765 | aF63.395985772412438
766 | aF63.304633119710324
767 | aF63.244056036892381
768 | aF62.863230195638025
769 | aF62.518910178152332
770 | aF63.83664024024678
771 | aF62.87236493015596
772 | aF61.90930325060733
773 | aF63.240754162766549
774 | aF63.286356177593866
775 | aF62.970759691911347
776 | aF63.843450886135848
777 | aF62.416727496042085
778 | aF62.812350836984706
779 | aF63.567045589899998
780 | aF62.567269108861382
781 | aF63.068908719449425
782 | aF63.494312994821421
783 | aF64.223676670532797
784 | aF62.213590091076668
785 | aF63.679595790404925
786 | aF63.52922398136608
787 | aF62.778823489729042
788 | aF62.624552231210103
789 | aF63.24203478192215
790 | aF62.9373737986988
791 | aF62.931055348865399
792 | aF63.070402994737506
793 | aF63.707584657985478
794 | aF63.645157451728423
795 | aF63.697964647644284
796 | aF62.978258152848262
797 | aF63.226834541459183
798 | aF62.784351611176618
799 | aF62.305742647244401
800 | aF62.224553870083547
801 | aF62.915963202983036
802 | aF63.033783611327799
803 | aF63.144720631682006
804 | aF63.965301138350334
805 | aF63.79793608193377
806 | aF63.610970033074317
807 | aF63.569829696523428
808 | aF63.174658618723079
809 | aF63.246340212829296
810 | aF62.450140691228853
811 | aF63.063228743355559
812 | aF63.53383196630508
813 | aF63.645729681715487
814 | aF63.73134412714942
815 | aF63.14078849074901
816 | aF63.484503462349132
817 | aF63.569303333316313
818 | aF62.916609651687921
819 | aF61.764062732878763
820 | aF61.843494503983443
821 | aF64.454409069720924
822 | aF62.983056352496384
823 | aF63.460213208144651
824 | aF63.562961884783512
825 | aF63.568536621971177
826 | aF63.373300974687389
827 | aF62.902163001578053
828 | aF63.252441705577695
829 | aF62.624911406531822
830 | aF63.805016731544015
831 | aF63.040901895569057
832 | aF63.112083395503674
833 | aF62.700907576257968
834 | aF61.883194221351459
835 | aF63.750022169951158
836 | aF62.56146614360874
837 | aF63.976536886119781
838 | aF63.035684595572235
839 | aF62.550905592351569
840 | aF62.852789136615741
841 | aF62.367665193649721
842 | aF62.71134121837494
843 | aF62.851377920975253
844 | aF63.157835258362958
845 | aF62.141492617649732
846 | aF63.814585267714499
847 | aF63.164351814465213
848 | aF63.428218356519992
849 | aF62.671660292057879
850 | aF63.416188632899662
851 | aF62.875119703353157
852 | aF62.38486792973557
853 | aF63.164498739511501
854 | aF61.987268121540374
855 | aF63.572530951084381
856 | aF63.085301504095803
857 | aF63.278164313783861
858 | aF63.005439020205323
859 | aF62.813403243740005
860 | aF63.191840039205651
861 | aF62.689019570768501
862 | aF62.831456305491002
863 | aF63.719449864125195
864 | aF63.300589854007093
865 | aF63.9059453204964
866 | aF63.030166793256399
867 | aF61.976361391686439
868 | aF62.4314742666635
869 | aF63.094630606648792
870 | aF62.59517735852446
871 | aF63.313357062055879
872 | aF63.691511088532238
873 | aF63.161190476847501
874 | aF64.030061578623958
875 | aF63.522228091447914
876 | aF63.164691999526362
877 | aF63.026579671510405
878 | aF62.962717087057293
879 | aF63.074618859197585
880 | aF63.473268485927122
881 | aF62.843057950519608
882 | aF64.074515122704071
883 | aF62.824349640930095
884 | aF62.674453001402163
885 | aF62.660314959620834
886 | aF63.378616282035104
887 | aF63.393369713293154
888 | aF63.403691992750652
889 | aF63.710427788645326
890 | aF63.371808935265605
891 | aF62.922077880175017
892 | aF63.784732773332422
893 | aF63.846036129109507
894 | aF62.361058057415967
895 | aF62.386778887271355
896 | aF63.395023855877433
897 | aF63.07063199256126
898 | aF63.80587407861811
899 | aF63.187788863124311
900 | aF65.592132541706263
901 | aF61.939024018745016
902 | aF62.530750579559772
903 | aF62.725981018732362
904 | aF62.790695833663456
905 | aF62.973090607040703
906 | aF63.512942082033923
907 | aF62.987170972030988
908 | aF63.508297721068338
909 | aF63.718495397770077
910 | aF62.582011863320837
911 | aF63.385676120225163
912 | aF63.634109079237099
913 | aF63.32002947317784
914 | aF62.810749338864625
915 | aF63.936519645662635
916 | aF62.677247203592934
917 | aF63.193167826680096
918 | aF63.220927445398168
919 | aF63.43972430453254
920 | aF63.57685816579184
921 | aF62.450336860907754
922 | aF63.585224580005672
923 | aF63.60465070472349
924 | aF64.424328736553363
925 | aF63.16409979988606
926 | aF63.476734619679434
927 | aF62.766138880903355
928 | aF63.614388641895232
929 | aF62.856892837817085
930 | aF62.947448146098978
931 | aF64.100349270520439
932 | aF62.875527363972154
933 | aF63.940606043702452
934 | aF63.234021687619595
935 | aF63.294628733370246
936 | aF63.029369490696617
937 | aF63.133857360254623
938 | aF63.847890358170616
939 | aF62.849773493946415
940 | aF62.882073772513976
941 | aF63.62765474679464
942 | aF62.313852528089427
943 | aF62.653850622032486
944 | aF62.599083016492514
945 | aF63.953732276759787
946 | aF62.883054841748454
947 | aF62.75166249118864
948 | aF62.614638632763658
949 | aF63.033061569236153
950 | aF62.898116211182156
951 | aF64.841656512483311
952 | aF65.371058282649685
953 | aF65.356456146192869
954 | aF65.640134591099894
955 | aF65.186427005777105
956 | aF65.372729231561621
957 | aF65.247842317453433
958 | aF65.757305569504553
959 | aF65.148663997847422
960 | aF64.885450065586269
961 | aF65.1736179004387
962 | aF65.522184444411991
963 | aF65.118771451761418
964 | aF64.967024363992664
965 | aF64.894148019359122
966 | aF65.379170598891591
967 | aF65.867999650822711
968 | aF65.262897453152448
969 | aF65.719230739526353
970 | aF64.758401260867259
971 | aF64.969914454301261
972 | a.
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | GNU GENERAL PUBLIC LICENSE
2 | Version 3, 29 June 2007
3 |
4 | Copyright (C) 2007 Free Software Foundation, Inc.
5 | Everyone is permitted to copy and distribute verbatim copies
6 | of this license document, but changing it is not allowed.
7 |
8 | Preamble
9 |
10 | The GNU General Public License is a free, copyleft license for
11 | software and other kinds of works.
12 |
13 | The licenses for most software and other practical works are designed
14 | to take away your freedom to share and change the works. By contrast,
15 | the GNU General Public License is intended to guarantee your freedom to
16 | share and change all versions of a program--to make sure it remains free
17 | software for all its users. We, the Free Software Foundation, use the
18 | GNU General Public License for most of our software; it applies also to
19 | any other work released this way by its authors. You can apply it to
20 | your programs, too.
21 |
22 | When we speak of free software, we are referring to freedom, not
23 | price. Our General Public Licenses are designed to make sure that you
24 | have the freedom to distribute copies of free software (and charge for
25 | them if you wish), that you receive source code or can get it if you
26 | want it, that you can change the software or use pieces of it in new
27 | free programs, and that you know you can do these things.
28 |
29 | To protect your rights, we need to prevent others from denying you
30 | these rights or asking you to surrender the rights. Therefore, you have
31 | certain responsibilities if you distribute copies of the software, or if
32 | you modify it: responsibilities to respect the freedom of others.
33 |
34 | For example, if you distribute copies of such a program, whether
35 | gratis or for a fee, you must pass on to the recipients the same
36 | freedoms that you received. You must make sure that they, too, receive
37 | or can get the source code. And you must show them these terms so they
38 | know their rights.
39 |
40 | Developers that use the GNU GPL protect your rights with two steps:
41 | (1) assert copyright on the software, and (2) offer you this License
42 | giving you legal permission to copy, distribute and/or modify it.
43 |
44 | For the developers' and authors' protection, the GPL clearly explains
45 | that there is no warranty for this free software. For both users' and
46 | authors' sake, the GPL requires that modified versions be marked as
47 | changed, so that their problems will not be attributed erroneously to
48 | authors of previous versions.
49 |
50 | Some devices are designed to deny users access to install or run
51 | modified versions of the software inside them, although the manufacturer
52 | can do so. This is fundamentally incompatible with the aim of
53 | protecting users' freedom to change the software. The systematic
54 | pattern of such abuse occurs in the area of products for individuals to
55 | use, which is precisely where it is most unacceptable. Therefore, we
56 | have designed this version of the GPL to prohibit the practice for those
57 | products. If such problems arise substantially in other domains, we
58 | stand ready to extend this provision to those domains in future versions
59 | of the GPL, as needed to protect the freedom of users.
60 |
61 | Finally, every program is threatened constantly by software patents.
62 | States should not allow patents to restrict development and use of
63 | software on general-purpose computers, but in those that do, we wish to
64 | avoid the special danger that patents applied to a free program could
65 | make it effectively proprietary. To prevent this, the GPL assures that
66 | patents cannot be used to render the program non-free.
67 |
68 | The precise terms and conditions for copying, distribution and
69 | modification follow.
70 |
71 | TERMS AND CONDITIONS
72 |
73 | 0. Definitions.
74 |
75 | "This License" refers to version 3 of the GNU General Public License.
76 |
77 | "Copyright" also means copyright-like laws that apply to other kinds of
78 | works, such as semiconductor masks.
79 |
80 | "The Program" refers to any copyrightable work licensed under this
81 | License. Each licensee is addressed as "you". "Licensees" and
82 | "recipients" may be individuals or organizations.
83 |
84 | To "modify" a work means to copy from or adapt all or part of the work
85 | in a fashion requiring copyright permission, other than the making of an
86 | exact copy. The resulting work is called a "modified version" of the
87 | earlier work or a work "based on" the earlier work.
88 |
89 | A "covered work" means either the unmodified Program or a work based
90 | on the Program.
91 |
92 | To "propagate" a work means to do anything with it that, without
93 | permission, would make you directly or secondarily liable for
94 | infringement under applicable copyright law, except executing it on a
95 | computer or modifying a private copy. Propagation includes copying,
96 | distribution (with or without modification), making available to the
97 | public, and in some countries other activities as well.
98 |
99 | To "convey" a work means any kind of propagation that enables other
100 | parties to make or receive copies. Mere interaction with a user through
101 | a computer network, with no transfer of a copy, is not conveying.
102 |
103 | An interactive user interface displays "Appropriate Legal Notices"
104 | to the extent that it includes a convenient and prominently visible
105 | feature that (1) displays an appropriate copyright notice, and (2)
106 | tells the user that there is no warranty for the work (except to the
107 | extent that warranties are provided), that licensees may convey the
108 | work under this License, and how to view a copy of this License. If
109 | the interface presents a list of user commands or options, such as a
110 | menu, a prominent item in the list meets this criterion.
111 |
112 | 1. Source Code.
113 |
114 | The "source code" for a work means the preferred form of the work
115 | for making modifications to it. "Object code" means any non-source
116 | form of a work.
117 |
118 | A "Standard Interface" means an interface that either is an official
119 | standard defined by a recognized standards body, or, in the case of
120 | interfaces specified for a particular programming language, one that
121 | is widely used among developers working in that language.
122 |
123 | The "System Libraries" of an executable work include anything, other
124 | than the work as a whole, that (a) is included in the normal form of
125 | packaging a Major Component, but which is not part of that Major
126 | Component, and (b) serves only to enable use of the work with that
127 | Major Component, or to implement a Standard Interface for which an
128 | implementation is available to the public in source code form. A
129 | "Major Component", in this context, means a major essential component
130 | (kernel, window system, and so on) of the specific operating system
131 | (if any) on which the executable work runs, or a compiler used to
132 | produce the work, or an object code interpreter used to run it.
133 |
134 | The "Corresponding Source" for a work in object code form means all
135 | the source code needed to generate, install, and (for an executable
136 | work) run the object code and to modify the work, including scripts to
137 | control those activities. However, it does not include the work's
138 | System Libraries, or general-purpose tools or generally available free
139 | programs which are used unmodified in performing those activities but
140 | which are not part of the work. For example, Corresponding Source
141 | includes interface definition files associated with source files for
142 | the work, and the source code for shared libraries and dynamically
143 | linked subprograms that the work is specifically designed to require,
144 | such as by intimate data communication or control flow between those
145 | subprograms and other parts of the work.
146 |
147 | The Corresponding Source need not include anything that users
148 | can regenerate automatically from other parts of the Corresponding
149 | Source.
150 |
151 | The Corresponding Source for a work in source code form is that
152 | same work.
153 |
154 | 2. Basic Permissions.
155 |
156 | All rights granted under this License are granted for the term of
157 | copyright on the Program, and are irrevocable provided the stated
158 | conditions are met. This License explicitly affirms your unlimited
159 | permission to run the unmodified Program. The output from running a
160 | covered work is covered by this License only if the output, given its
161 | content, constitutes a covered work. This License acknowledges your
162 | rights of fair use or other equivalent, as provided by copyright law.
163 |
164 | You may make, run and propagate covered works that you do not
165 | convey, without conditions so long as your license otherwise remains
166 | in force. You may convey covered works to others for the sole purpose
167 | of having them make modifications exclusively for you, or provide you
168 | with facilities for running those works, provided that you comply with
169 | the terms of this License in conveying all material for which you do
170 | not control copyright. Those thus making or running the covered works
171 | for you must do so exclusively on your behalf, under your direction
172 | and control, on terms that prohibit them from making any copies of
173 | your copyrighted material outside their relationship with you.
174 |
175 | Conveying under any other circumstances is permitted solely under
176 | the conditions stated below. Sublicensing is not allowed; section 10
177 | makes it unnecessary.
178 |
179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
180 |
181 | No covered work shall be deemed part of an effective technological
182 | measure under any applicable law fulfilling obligations under article
183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or
184 | similar laws prohibiting or restricting circumvention of such
185 | measures.
186 |
187 | When you convey a covered work, you waive any legal power to forbid
188 | circumvention of technological measures to the extent such circumvention
189 | is effected by exercising rights under this License with respect to
190 | the covered work, and you disclaim any intention to limit operation or
191 | modification of the work as a means of enforcing, against the work's
192 | users, your or third parties' legal rights to forbid circumvention of
193 | technological measures.
194 |
195 | 4. Conveying Verbatim Copies.
196 |
197 | You may convey verbatim copies of the Program's source code as you
198 | receive it, in any medium, provided that you conspicuously and
199 | appropriately publish on each copy an appropriate copyright notice;
200 | keep intact all notices stating that this License and any
201 | non-permissive terms added in accord with section 7 apply to the code;
202 | keep intact all notices of the absence of any warranty; and give all
203 | recipients a copy of this License along with the Program.
204 |
205 | You may charge any price or no price for each copy that you convey,
206 | and you may offer support or warranty protection for a fee.
207 |
208 | 5. Conveying Modified Source Versions.
209 |
210 | You may convey a work based on the Program, or the modifications to
211 | produce it from the Program, in the form of source code under the
212 | terms of section 4, provided that you also meet all of these conditions:
213 |
214 | a) The work must carry prominent notices stating that you modified
215 | it, and giving a relevant date.
216 |
217 | b) The work must carry prominent notices stating that it is
218 | released under this License and any conditions added under section
219 | 7. This requirement modifies the requirement in section 4 to
220 | "keep intact all notices".
221 |
222 | c) You must license the entire work, as a whole, under this
223 | License to anyone who comes into possession of a copy. This
224 | License will therefore apply, along with any applicable section 7
225 | additional terms, to the whole of the work, and all its parts,
226 | regardless of how they are packaged. This License gives no
227 | permission to license the work in any other way, but it does not
228 | invalidate such permission if you have separately received it.
229 |
230 | d) If the work has interactive user interfaces, each must display
231 | Appropriate Legal Notices; however, if the Program has interactive
232 | interfaces that do not display Appropriate Legal Notices, your
233 | work need not make them do so.
234 |
235 | A compilation of a covered work with other separate and independent
236 | works, which are not by their nature extensions of the covered work,
237 | and which are not combined with it such as to form a larger program,
238 | in or on a volume of a storage or distribution medium, is called an
239 | "aggregate" if the compilation and its resulting copyright are not
240 | used to limit the access or legal rights of the compilation's users
241 | beyond what the individual works permit. Inclusion of a covered work
242 | in an aggregate does not cause this License to apply to the other
243 | parts of the aggregate.
244 |
245 | 6. Conveying Non-Source Forms.
246 |
247 | You may convey a covered work in object code form under the terms
248 | of sections 4 and 5, provided that you also convey the
249 | machine-readable Corresponding Source under the terms of this License,
250 | in one of these ways:
251 |
252 | a) Convey the object code in, or embodied in, a physical product
253 | (including a physical distribution medium), accompanied by the
254 | Corresponding Source fixed on a durable physical medium
255 | customarily used for software interchange.
256 |
257 | b) Convey the object code in, or embodied in, a physical product
258 | (including a physical distribution medium), accompanied by a
259 | written offer, valid for at least three years and valid for as
260 | long as you offer spare parts or customer support for that product
261 | model, to give anyone who possesses the object code either (1) a
262 | copy of the Corresponding Source for all the software in the
263 | product that is covered by this License, on a durable physical
264 | medium customarily used for software interchange, for a price no
265 | more than your reasonable cost of physically performing this
266 | conveying of source, or (2) access to copy the
267 | Corresponding Source from a network server at no charge.
268 |
269 | c) Convey individual copies of the object code with a copy of the
270 | written offer to provide the Corresponding Source. This
271 | alternative is allowed only occasionally and noncommercially, and
272 | only if you received the object code with such an offer, in accord
273 | with subsection 6b.
274 |
275 | d) Convey the object code by offering access from a designated
276 | place (gratis or for a charge), and offer equivalent access to the
277 | Corresponding Source in the same way through the same place at no
278 | further charge. You need not require recipients to copy the
279 | Corresponding Source along with the object code. If the place to
280 | copy the object code is a network server, the Corresponding Source
281 | may be on a different server (operated by you or a third party)
282 | that supports equivalent copying facilities, provided you maintain
283 | clear directions next to the object code saying where to find the
284 | Corresponding Source. Regardless of what server hosts the
285 | Corresponding Source, you remain obligated to ensure that it is
286 | available for as long as needed to satisfy these requirements.
287 |
288 | e) Convey the object code using peer-to-peer transmission, provided
289 | you inform other peers where the object code and Corresponding
290 | Source of the work are being offered to the general public at no
291 | charge under subsection 6d.
292 |
293 | A separable portion of the object code, whose source code is excluded
294 | from the Corresponding Source as a System Library, need not be
295 | included in conveying the object code work.
296 |
297 | A "User Product" is either (1) a "consumer product", which means any
298 | tangible personal property which is normally used for personal, family,
299 | or household purposes, or (2) anything designed or sold for incorporation
300 | into a dwelling. In determining whether a product is a consumer product,
301 | doubtful cases shall be resolved in favor of coverage. For a particular
302 | product received by a particular user, "normally used" refers to a
303 | typical or common use of that class of product, regardless of the status
304 | of the particular user or of the way in which the particular user
305 | actually uses, or expects or is expected to use, the product. A product
306 | is a consumer product regardless of whether the product has substantial
307 | commercial, industrial or non-consumer uses, unless such uses represent
308 | the only significant mode of use of the product.
309 |
310 | "Installation Information" for a User Product means any methods,
311 | procedures, authorization keys, or other information required to install
312 | and execute modified versions of a covered work in that User Product from
313 | a modified version of its Corresponding Source. The information must
314 | suffice to ensure that the continued functioning of the modified object
315 | code is in no case prevented or interfered with solely because
316 | modification has been made.
317 |
318 | If you convey an object code work under this section in, or with, or
319 | specifically for use in, a User Product, and the conveying occurs as
320 | part of a transaction in which the right of possession and use of the
321 | User Product is transferred to the recipient in perpetuity or for a
322 | fixed term (regardless of how the transaction is characterized), the
323 | Corresponding Source conveyed under this section must be accompanied
324 | by the Installation Information. But this requirement does not apply
325 | if neither you nor any third party retains the ability to install
326 | modified object code on the User Product (for example, the work has
327 | been installed in ROM).
328 |
329 | The requirement to provide Installation Information does not include a
330 | requirement to continue to provide support service, warranty, or updates
331 | for a work that has been modified or installed by the recipient, or for
332 | the User Product in which it has been modified or installed. Access to a
333 | network may be denied when the modification itself materially and
334 | adversely affects the operation of the network or violates the rules and
335 | protocols for communication across the network.
336 |
337 | Corresponding Source conveyed, and Installation Information provided,
338 | in accord with this section must be in a format that is publicly
339 | documented (and with an implementation available to the public in
340 | source code form), and must require no special password or key for
341 | unpacking, reading or copying.
342 |
343 | 7. Additional Terms.
344 |
345 | "Additional permissions" are terms that supplement the terms of this
346 | License by making exceptions from one or more of its conditions.
347 | Additional permissions that are applicable to the entire Program shall
348 | be treated as though they were included in this License, to the extent
349 | that they are valid under applicable law. If additional permissions
350 | apply only to part of the Program, that part may be used separately
351 | under those permissions, but the entire Program remains governed by
352 | this License without regard to the additional permissions.
353 |
354 | When you convey a copy of a covered work, you may at your option
355 | remove any additional permissions from that copy, or from any part of
356 | it. (Additional permissions may be written to require their own
357 | removal in certain cases when you modify the work.) You may place
358 | additional permissions on material, added by you to a covered work,
359 | for which you have or can give appropriate copyright permission.
360 |
361 | Notwithstanding any other provision of this License, for material you
362 | add to a covered work, you may (if authorized by the copyright holders of
363 | that material) supplement the terms of this License with terms:
364 |
365 | a) Disclaiming warranty or limiting liability differently from the
366 | terms of sections 15 and 16 of this License; or
367 |
368 | b) Requiring preservation of specified reasonable legal notices or
369 | author attributions in that material or in the Appropriate Legal
370 | Notices displayed by works containing it; or
371 |
372 | c) Prohibiting misrepresentation of the origin of that material, or
373 | requiring that modified versions of such material be marked in
374 | reasonable ways as different from the original version; or
375 |
376 | d) Limiting the use for publicity purposes of names of licensors or
377 | authors of the material; or
378 |
379 | e) Declining to grant rights under trademark law for use of some
380 | trade names, trademarks, or service marks; or
381 |
382 | f) Requiring indemnification of licensors and authors of that
383 | material by anyone who conveys the material (or modified versions of
384 | it) with contractual assumptions of liability to the recipient, for
385 | any liability that these contractual assumptions directly impose on
386 | those licensors and authors.
387 |
388 | All other non-permissive additional terms are considered "further
389 | restrictions" within the meaning of section 10. If the Program as you
390 | received it, or any part of it, contains a notice stating that it is
391 | governed by this License along with a term that is a further
392 | restriction, you may remove that term. If a license document contains
393 | a further restriction but permits relicensing or conveying under this
394 | License, you may add to a covered work material governed by the terms
395 | of that license document, provided that the further restriction does
396 | not survive such relicensing or conveying.
397 |
398 | If you add terms to a covered work in accord with this section, you
399 | must place, in the relevant source files, a statement of the
400 | additional terms that apply to those files, or a notice indicating
401 | where to find the applicable terms.
402 |
403 | Additional terms, permissive or non-permissive, may be stated in the
404 | form of a separately written license, or stated as exceptions;
405 | the above requirements apply either way.
406 |
407 | 8. Termination.
408 |
409 | You may not propagate or modify a covered work except as expressly
410 | provided under this License. Any attempt otherwise to propagate or
411 | modify it is void, and will automatically terminate your rights under
412 | this License (including any patent licenses granted under the third
413 | paragraph of section 11).
414 |
415 | However, if you cease all violation of this License, then your
416 | license from a particular copyright holder is reinstated (a)
417 | provisionally, unless and until the copyright holder explicitly and
418 | finally terminates your license, and (b) permanently, if the copyright
419 | holder fails to notify you of the violation by some reasonable means
420 | prior to 60 days after the cessation.
421 |
422 | Moreover, your license from a particular copyright holder is
423 | reinstated permanently if the copyright holder notifies you of the
424 | violation by some reasonable means, this is the first time you have
425 | received notice of violation of this License (for any work) from that
426 | copyright holder, and you cure the violation prior to 30 days after
427 | your receipt of the notice.
428 |
429 | Termination of your rights under this section does not terminate the
430 | licenses of parties who have received copies or rights from you under
431 | this License. If your rights have been terminated and not permanently
432 | reinstated, you do not qualify to receive new licenses for the same
433 | material under section 10.
434 |
435 | 9. Acceptance Not Required for Having Copies.
436 |
437 | You are not required to accept this License in order to receive or
438 | run a copy of the Program. Ancillary propagation of a covered work
439 | occurring solely as a consequence of using peer-to-peer transmission
440 | to receive a copy likewise does not require acceptance. However,
441 | nothing other than this License grants you permission to propagate or
442 | modify any covered work. These actions infringe copyright if you do
443 | not accept this License. Therefore, by modifying or propagating a
444 | covered work, you indicate your acceptance of this License to do so.
445 |
446 | 10. Automatic Licensing of Downstream Recipients.
447 |
448 | Each time you convey a covered work, the recipient automatically
449 | receives a license from the original licensors, to run, modify and
450 | propagate that work, subject to this License. You are not responsible
451 | for enforcing compliance by third parties with this License.
452 |
453 | An "entity transaction" is a transaction transferring control of an
454 | organization, or substantially all assets of one, or subdividing an
455 | organization, or merging organizations. If propagation of a covered
456 | work results from an entity transaction, each party to that
457 | transaction who receives a copy of the work also receives whatever
458 | licenses to the work the party's predecessor in interest had or could
459 | give under the previous paragraph, plus a right to possession of the
460 | Corresponding Source of the work from the predecessor in interest, if
461 | the predecessor has it or can get it with reasonable efforts.
462 |
463 | You may not impose any further restrictions on the exercise of the
464 | rights granted or affirmed under this License. For example, you may
465 | not impose a license fee, royalty, or other charge for exercise of
466 | rights granted under this License, and you may not initiate litigation
467 | (including a cross-claim or counterclaim in a lawsuit) alleging that
468 | any patent claim is infringed by making, using, selling, offering for
469 | sale, or importing the Program or any portion of it.
470 |
471 | 11. Patents.
472 |
473 | A "contributor" is a copyright holder who authorizes use under this
474 | License of the Program or a work on which the Program is based. The
475 | work thus licensed is called the contributor's "contributor version".
476 |
477 | A contributor's "essential patent claims" are all patent claims
478 | owned or controlled by the contributor, whether already acquired or
479 | hereafter acquired, that would be infringed by some manner, permitted
480 | by this License, of making, using, or selling its contributor version,
481 | but do not include claims that would be infringed only as a
482 | consequence of further modification of the contributor version. For
483 | purposes of this definition, "control" includes the right to grant
484 | patent sublicenses in a manner consistent with the requirements of
485 | this License.
486 |
487 | Each contributor grants you a non-exclusive, worldwide, royalty-free
488 | patent license under the contributor's essential patent claims, to
489 | make, use, sell, offer for sale, import and otherwise run, modify and
490 | propagate the contents of its contributor version.
491 |
492 | In the following three paragraphs, a "patent license" is any express
493 | agreement or commitment, however denominated, not to enforce a patent
494 | (such as an express permission to practice a patent or covenant not to
495 | sue for patent infringement). To "grant" such a patent license to a
496 | party means to make such an agreement or commitment not to enforce a
497 | patent against the party.
498 |
499 | If you convey a covered work, knowingly relying on a patent license,
500 | and the Corresponding Source of the work is not available for anyone
501 | to copy, free of charge and under the terms of this License, through a
502 | publicly available network server or other readily accessible means,
503 | then you must either (1) cause the Corresponding Source to be so
504 | available, or (2) arrange to deprive yourself of the benefit of the
505 | patent license for this particular work, or (3) arrange, in a manner
506 | consistent with the requirements of this License, to extend the patent
507 | license to downstream recipients. "Knowingly relying" means you have
508 | actual knowledge that, but for the patent license, your conveying the
509 | covered work in a country, or your recipient's use of the covered work
510 | in a country, would infringe one or more identifiable patents in that
511 | country that you have reason to believe are valid.
512 |
513 | If, pursuant to or in connection with a single transaction or
514 | arrangement, you convey, or propagate by procuring conveyance of, a
515 | covered work, and grant a patent license to some of the parties
516 | receiving the covered work authorizing them to use, propagate, modify
517 | or convey a specific copy of the covered work, then the patent license
518 | you grant is automatically extended to all recipients of the covered
519 | work and works based on it.
520 |
521 | A patent license is "discriminatory" if it does not include within
522 | the scope of its coverage, prohibits the exercise of, or is
523 | conditioned on the non-exercise of one or more of the rights that are
524 | specifically granted under this License. You may not convey a covered
525 | work if you are a party to an arrangement with a third party that is
526 | in the business of distributing software, under which you make payment
527 | to the third party based on the extent of your activity of conveying
528 | the work, and under which the third party grants, to any of the
529 | parties who would receive the covered work from you, a discriminatory
530 | patent license (a) in connection with copies of the covered work
531 | conveyed by you (or copies made from those copies), or (b) primarily
532 | for and in connection with specific products or compilations that
533 | contain the covered work, unless you entered into that arrangement,
534 | or that patent license was granted, prior to 28 March 2007.
535 |
536 | Nothing in this License shall be construed as excluding or limiting
537 | any implied license or other defenses to infringement that may
538 | otherwise be available to you under applicable patent law.
539 |
540 | 12. No Surrender of Others' Freedom.
541 |
542 | If conditions are imposed on you (whether by court order, agreement or
543 | otherwise) that contradict the conditions of this License, they do not
544 | excuse you from the conditions of this License. If you cannot convey a
545 | covered work so as to satisfy simultaneously your obligations under this
546 | License and any other pertinent obligations, then as a consequence you may
547 | not convey it at all. For example, if you agree to terms that obligate you
548 | to collect a royalty for further conveying from those to whom you convey
549 | the Program, the only way you could satisfy both those terms and this
550 | License would be to refrain entirely from conveying the Program.
551 |
552 | 13. Use with the GNU Affero General Public License.
553 |
554 | Notwithstanding any other provision of this License, you have
555 | permission to link or combine any covered work with a work licensed
556 | under version 3 of the GNU Affero General Public License into a single
557 | combined work, and to convey the resulting work. The terms of this
558 | License will continue to apply to the part which is the covered work,
559 | but the special requirements of the GNU Affero General Public License,
560 | section 13, concerning interaction through a network will apply to the
561 | combination as such.
562 |
563 | 14. Revised Versions of this License.
564 |
565 | The Free Software Foundation may publish revised and/or new versions of
566 | the GNU General Public License from time to time. Such new versions will
567 | be similar in spirit to the present version, but may differ in detail to
568 | address new problems or concerns.
569 |
570 | Each version is given a distinguishing version number. If the
571 | Program specifies that a certain numbered version of the GNU General
572 | Public License "or any later version" applies to it, you have the
573 | option of following the terms and conditions either of that numbered
574 | version or of any later version published by the Free Software
575 | Foundation. If the Program does not specify a version number of the
576 | GNU General Public License, you may choose any version ever published
577 | by the Free Software Foundation.
578 |
579 | If the Program specifies that a proxy can decide which future
580 | versions of the GNU General Public License can be used, that proxy's
581 | public statement of acceptance of a version permanently authorizes you
582 | to choose that version for the Program.
583 |
584 | Later license versions may give you additional or different
585 | permissions. However, no additional obligations are imposed on any
586 | author or copyright holder as a result of your choosing to follow a
587 | later version.
588 |
589 | 15. Disclaimer of Warranty.
590 |
591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
599 |
600 | 16. Limitation of Liability.
601 |
602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
610 | SUCH DAMAGES.
611 |
612 | 17. Interpretation of Sections 15 and 16.
613 |
614 | If the disclaimer of warranty and limitation of liability provided
615 | above cannot be given local legal effect according to their terms,
616 | reviewing courts shall apply local law that most closely approximates
617 | an absolute waiver of all civil liability in connection with the
618 | Program, unless a warranty or assumption of liability accompanies a
619 | copy of the Program in return for a fee.
620 |
621 | END OF TERMS AND CONDITIONS
622 |
623 | How to Apply These Terms to Your New Programs
624 |
625 | If you develop a new program, and you want it to be of the greatest
626 | possible use to the public, the best way to achieve this is to make it
627 | free software which everyone can redistribute and change under these terms.
628 |
629 | To do so, attach the following notices to the program. It is safest
630 | to attach them to the start of each source file to most effectively
631 | state the exclusion of warranty; and each file should have at least
632 | the "copyright" line and a pointer to where the full notice is found.
633 |
634 | {one line to give the program's name and a brief idea of what it does.}
635 | Copyright (C) {year} {name of author}
636 |
637 | This program is free software: you can redistribute it and/or modify
638 | it under the terms of the GNU General Public License as published by
639 | the Free Software Foundation, either version 3 of the License, or
640 | (at your option) any later version.
641 |
642 | This program is distributed in the hope that it will be useful,
643 | but WITHOUT ANY WARRANTY; without even the implied warranty of
644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
645 | GNU General Public License for more details.
646 |
647 | You should have received a copy of the GNU General Public License
648 | along with this program. If not, see .
649 |
650 | Also add information on how to contact you by electronic and paper mail.
651 |
652 | If the program does terminal interaction, make it output a short
653 | notice like this when it starts in an interactive mode:
654 |
655 | {project} Copyright (C) {year} {fullname}
656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
657 | This is free software, and you are welcome to redistribute it
658 | under certain conditions; type `show c' for details.
659 |
660 | The hypothetical commands `show w' and `show c' should show the appropriate
661 | parts of the General Public License. Of course, your program's commands
662 | might be different; for a GUI interface, you would use an "about box".
663 |
664 | You should also get your employer (if you work as a programmer) or school,
665 | if any, to sign a "copyright disclaimer" for the program, if necessary.
666 | For more information on this, and how to apply and follow the GNU GPL, see
667 | .
668 |
669 | The GNU General Public License does not permit incorporating your program
670 | into proprietary programs. If your program is a subroutine library, you
671 | may consider it more useful to permit linking proprietary applications with
672 | the library. If this is what you want to do, use the GNU Lesser General
673 | Public License instead of this License. But first, please read
674 | .
675 |
--------------------------------------------------------------------------------
/examples/Confusion_matrix.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "## Confusion matrix for unfiltered trajectories.\n",
8 | "This notebook computes the confusion matrix for unfiltered synthetic data for several thresholds. When the data is unfiltered, many change points are missed. However, as the threshold is lowered and the false negatives decrease, the false positive rate increases. "
9 | ]
10 | },
11 | {
12 | "cell_type": "code",
13 | "execution_count": 1,
14 | "metadata": {
15 | "collapsed": true
16 | },
17 | "outputs": [],
18 | "source": [
19 | "%matplotlib inline\n",
20 | "import matplotlib.pyplot as plt\n",
21 | "import matplotlib as mpl\n",
22 | "import numpy as np\n",
23 | "from cpdetect import cpDetector\n",
24 | "import pandas as pd\n",
25 | "from matplotlib.backends.backend_pdf import PdfPages\n",
26 | "from tqdm import *\n",
27 | "try:\n",
28 | " import cPickle as pickle\n",
29 | "except:\n",
30 | " import pickle\n",
31 | "import os, glob"
32 | ]
33 | },
34 | {
35 | "cell_type": "code",
36 | "execution_count": 2,
37 | "metadata": {
38 | "collapsed": true
39 | },
40 | "outputs": [],
41 | "source": [
42 | "# Load all relavant data (synthetic trajectories and pickled detectors.)\n",
43 | "trajs = np.load('synthetic_trajs.np.npy')\n",
44 | "filtered_trajs = np.load('filtered_trajs.npy')\n",
45 | "true_ts = pickle.load(open('true_ts.pickle', 'rb'))\n",
46 | "true_step = pickle.load(open('step_synthetic.pickle', 'rb'))\n",
47 | "files = [file for file in glob.glob('detector*')]\n",
48 | "detectors = {f[9:-7]: pickle.load(open(f, 'rb')) for f in files}"
49 | ]
50 | },
51 | {
52 | "cell_type": "code",
53 | "execution_count": 4,
54 | "metadata": {
55 | "collapsed": false
56 | },
57 | "outputs": [],
58 | "source": [
59 | "# Calculate all confusion matrices. The rejection window here is 100 timepoints. \n",
60 | "cm = np.zeros((10, 2, 2))\n",
61 | "for d in detectors:\n",
62 | " tp = 0\n",
63 | " fp = 0\n",
64 | " fn = 0\n",
65 | " for t in range(len(trajs)):\n",
66 | " true_positive = []\n",
67 | " false_negative = true_ts['traj_{}'.format(t)][:-1]\n",
68 | " false_positive = np.asarray(detectors[d].change_points['traj_{}'.format(t)]['ts'])\n",
69 | " index_neg = []\n",
70 | " for i, t_ts in enumerate(false_negative):\n",
71 | " for j, p_ts in enumerate(false_positive):\n",
72 | " if t_ts-100 <= p_ts <= t_ts+100:\n",
73 | " true_positive.append(p_ts)\n",
74 | " index_neg.append(i)\n",
75 | " false_positive = np.delete(false_positive, j)\n",
76 | " break\n",
77 | " false_negative = np.delete(false_negative, index_neg)\n",
78 | " # sanity check\n",
79 | " assert(len(true_ts['traj_{}'.format(t)][:-1]) == (len(true_positive) + len(false_negative)))\n",
80 | " assert(len(np.asarray(detectors[d].change_points['traj_{}'.format(t)]['ts']+28)) == \n",
81 | " (len(true_positive) + len(false_positive)))\n",
82 | " tp += len(true_positive)\n",
83 | " fp += len(false_positive)\n",
84 | " fn += len(false_negative)\n",
85 | " m = int(d[-1])\n",
86 | " cm[m] = np.array(([tp, fp], [fn, 0]))"
87 | ]
88 | },
89 | {
90 | "cell_type": "code",
91 | "execution_count": 5,
92 | "metadata": {
93 | "collapsed": false
94 | },
95 | "outputs": [
96 | {
97 | "data": {
98 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAAEKCAYAAADzQPVvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZO+ccyHgyd4550LAk71zzoXA/wddoYROFdQ5kAAAAABJRU5ErkJggg==\n",
99 | "text/plain": [
100 | ""
101 | ]
102 | },
103 | "metadata": {},
104 | "output_type": "display_data"
105 | },
106 | {
107 | "data": {
108 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAAEKCAYAAADzQPVvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
109 | "text/plain": [
110 | ""
111 | ]
112 | },
113 | "metadata": {},
114 | "output_type": "display_data"
115 | },
116 | {
117 | "data": {
118 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXsAAACJCAYAAADNE5itAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
119 | "text/plain": [
120 | ""
121 | ]
122 | },
123 | "metadata": {},
124 | "output_type": "display_data"
125 | }
126 | ],
127 | "source": [
128 | "# Plot confusion matrices\n",
129 | "key = [['TP', 'FN'], ['FP', 'TN']]\n",
130 | "labelx = ['True T', 'No T']\n",
131 | "labely = ['Predicted T', 'Predicted No T']\n",
132 | "filename = 'confusion_matrix.pdf'\n",
133 | "with PdfPages(filename) as pdf:\n",
134 | " for p in range(3):\n",
135 | " fig = plt.figure()\n",
136 | " for f in range(4):\n",
137 | " if p*4 + f > 9:\n",
138 | " break\n",
139 | " ax = fig.add_subplot(2, 2, f+1)\n",
140 | " confusion = ax.matshow(cm[p*4 + f], cmap='Blues')\n",
141 | " for (j, k), label in np.ndenumerate(cm[p*4+f]):\n",
142 | " l = key[k][j] + ': ' + str(int(label))\n",
143 | " ax.text(k,j,l,ha='center',va='center')\n",
144 | " plt.title('Threshold = -{}'.format(str(p*4+f)));\n",
145 | " if f in (2, 3):\n",
146 | " plt.xlabel('True')\n",
147 | " ax.set_xticklabels([])\n",
148 | " ax.set_yticklabels(['']+labely)\n",
149 | " pdf.savefig(bbox_inches='tight')\n",
150 | " "
151 | ]
152 | },
153 | {
154 | "cell_type": "code",
155 | "execution_count": null,
156 | "metadata": {
157 | "collapsed": true
158 | },
159 | "outputs": [],
160 | "source": []
161 | }
162 | ],
163 | "metadata": {
164 | "kernelspec": {
165 | "display_name": "Python 3",
166 | "language": "python",
167 | "name": "python3"
168 | },
169 | "language_info": {
170 | "codemirror_mode": {
171 | "name": "ipython",
172 | "version": 3
173 | },
174 | "file_extension": ".py",
175 | "mimetype": "text/x-python",
176 | "name": "python",
177 | "nbconvert_exporter": "python",
178 | "pygments_lexer": "ipython3",
179 | "version": "3.5.1"
180 | }
181 | },
182 | "nbformat": 4,
183 | "nbformat_minor": 0
184 | }
185 |
--------------------------------------------------------------------------------