├── README.md
└── CrowSearchAlgorithm.py


/README.md:
--------------------------------------------------------------------------------
 1 | # CrowSearchAlgorithmPython
 2 | A Python implementation of Crow Search Algorithm
 3 | 
 4 | ## Definition
 5 | 
 6 | Crow Search Algorithm is a search algorithm proposed by Alireza Askarzadeh, in the year 2016, as a new metaheuristic method aimed at solution of the most varied engineering optimization problems. Askarzadeh based his idea on the behavior of crows to develop the algorithm. Therefore this can be classified as bioinspired (inspired by nature).
 7 | 
 8 | A crow is an animal known to collect its food and when necessary to hide this food so that other predators do not have the opportunity to steal your food. This feature is developed by the fact that they watch other birds and where these birds store their stock of food, and as soon as the animals are no longer around, the crows come to the place and steal the food. Therefore, because they have these characteristics, the crows they understand the need to hide their food. 
 9 | 
10 | In this sense, [Askarzadeh (2016)][1] defines 4 principles that form the basis of thinking for the construction of the algorithm:
11 | 
12 | - Crows live in flocks;
13 | - Crows memorize the position of their hiding place;
14 | - Crows follow each other to steal food;
15 | - Crows protect your stock (by a probability).
16 | 
17 | ## References
18 | 
19 | [ASKARZADEH, Alireza. A novel metaheuristic method for solving constrained
20 | engineering optimization problems: Crow search algorithm. Computers &
21 | Structures, v. 169, p. 1-12, 2016.][1]
22 | 
23 | [1]: https://www.sciencedirect.com/science/article/abs/pii/S0045794916300475
24 | 
25 | 


--------------------------------------------------------------------------------
/CrowSearchAlgorithm.py:
--------------------------------------------------------------------------------
 1 | # -------------------------------------------------
 2 | # Citation details:
 3 | # Alireza Askarzadeh, Anovel metaheuristic method for solving constrained
 4 | # engineering optimization problems: Crow search algorithm, Computers &
 5 | # Structures, Vol. 169, 1-12, 2016.
 6 | # Programmed by Alireza Askarzadeh at Kerman Graduate #
 7 | # University of Advanced Technology(KGUT) #
 8 | # Date of programming: September 2015 #
 9 | # -------------------------------------------------
10 | # This demo only implements a standard version of CSA for minimization of
11 | # a standard test function(Sphere) on MATLAB 7.6.0 (R2008a).
12 | # -------------------------------------------------
13 | # Note:
14 | # Due to the stochastic nature of meta-heuristc algorithms, different runs
15 | # may lead to slightly different results.
16 | # -------------------------------------------------
17 | # Simple transcripted to SciLab by Luan Michel(github.com/Luan-Michel/CrowSearchAlgorithmPython)
18 | # Original in https: # www.mathworks.com/matlabcentral/fileexchange/56127-crow-search-algorithm
19 | 
20 | import random	#random Function
21 | import numpy 	#numpy operations
22 | import math		#ceil function
23 | 
24 | def init(n, pd, l, u): #init the matrix problem
25 | 	x = []
26 | 	for i in range (n):
27 | 		x.append([])
28 | 		for j in range (pd):
29 | 			x[i].append(l-(l-u)*(random.random()))
30 | 	return x
31 | 
32 | def fitness(xn, n ,pd):	#function for fitness calculation
33 |     fitness = []
34 |     for i in range(n):
35 |         fitness.append(0)
36 |         for j in range(pd):
37 |             fitness[i] = fitness[i]+pow(xn[(i, j)], 2)
38 |     return fitness
39 | 
40 | # variables initialization #
41 | pd = 10		#Problem dimension (number of decision variables)
42 | n = 20		#Flock (population) size
43 | ap = 0.1	#Awareness probability
44 | fl = 2		#Flight length (fl)
45 | l = -100	#Lower
46 | u = 100		#Uper
47 | x = numpy.matrix(init(n, pd, l, u))
48 | xn = x.copy()
49 | ft = numpy.array(fitness(xn, n, pd))
50 | mem = x.copy()			#Memory initialization
51 | fit_mem = ft.copy()		#Fitness of memory positions
52 | tmax = 1000				#Max numuber of iterations (itermax)
53 | ffit = numpy.array([])	# Best fit of each iteration
54 | 
55 | #Iteration begin
56 | 
57 | for t in range(tmax):
58 | 
59 | 	num = numpy.array([random.randint(0, n-1) for _ in range(n)]) # Generation of random candidate crows for following (chasing)
60 | 	xnew = numpy.empty((n,pd))
61 | 	for i in range (n):
62 | 		if(random.random() > ap):
63 | 			for j in range (pd):
64 | 				xnew[(i,j)] = x[(i,j)]+fl*((random.random())*(mem[(num[i],j)]-x[(i,j)]))
65 | 		else:
66 | 			for j in range (pd):
67 | 				xnew[(i, j)] = l-(l-u)*random.random()
68 | 	xn = xnew.copy()
69 | 	ft = numpy.array(fitness(xn, n, pd)) #Function for fitness evaluation of new solutions
70 | 
71 | 	#Update position and memory#
72 | 	for i in range(n):
73 | 		if(xnew[i].all() > l and xnew[i].all() < u):
74 | 			x[i] = xnew[i].copy()		#Update position
75 | 			if(ft[i] < fit_mem[i]):
76 | 				mem[i] = xnew[i].copy()	#Update memory
77 | 				fit_mem[i] = ft[i]
78 | 	ffit = numpy.append(ffit, numpy.amin(fit_mem)) 	#Best found value until iteration t
79 | 
80 | ngbest, = numpy.where(numpy.isclose(fit_mem, min(fit_mem)))
81 | print(mem[ngbest[0]])
82 | 


--------------------------------------------------------------------------------