├── ArtificialNeuralNetwork.py
├── ArtificialNeuralNetworkUtility.py
└── README.md


/ArtificialNeuralNetwork.py:
--------------------------------------------------------------------------------
 1 | from ArtificialNeuralNetworkUtility import sigmoid
 2 | from ArtificialNeuralNetworkUtility import between
 3 | from ArtificialNeuralNetworkUtility import make_matrix
 4 | 
 5 | use_bias = 1
 6 | squash = sigmoid
 7 | 
 8 | class ArtificialNeuralNetwork:
 9 | 
10 | 	def __init__(self, layer_sizes):
11 | 		self.layers = []
12 | 		self.learn_rate = 0.1
13 | 		
14 | 		for l in range(len(layer_sizes)):
15 | 			layer_size = layer_sizes[l]
16 | 			prev_layer_size = (0 if l == 0 else layer_sizes[l-1])
17 | 			layer = Layer(l, layer_size, prev_layer_size)
18 | 			self.layers.append(layer)
19 | 
20 | 	def train(self):
21 | 		pass
22 | 		
23 | 	def predict(self, input):
24 | 		"""
25 | 		Return the network prediction for the input.
26 | 		"""
27 | 		self.set_input(input)
28 | 		self.fwd_propagate()
29 | 		
30 | 		return self.get_output()
31 | 		
32 | 	def update_weights(self):
33 | 		pass
34 | 		
35 | 	def set_input(self, input_vector):
36 | 		input_layer = self.layers[0]
37 | 		
38 | 		for i in range(0, input_layer.n_neurons):
39 | 			input_layer.output[i + use_bias] = input_vector[i]
40 | 		
41 | 	def fwd_propagate(self):
42 | 		"""
43 | 		Forward propagate the input signal through the neural network.
44 | 		"""
45 | 		# Exclude the last layer
46 | 		for l in range(len(self.layers) - 1):
47 | 			src_layer = self.layers[l]
48 | 			dest_layer = self.layers[l + 1]
49 | 			for j in range(0, dest_layer.n_neurons):
50 | 				input_sum = 0
51 | 				for i in range(0, src_layer.n_neurons + use_bias):
52 | 					input_sum += dest_layer.weight[i][j] * src_layer.output[i]
53 | 				dest_layer.input[j] = input_sum
54 | 				dest_layer.output[j + use_bias] = squash(input_sum)
55 | 				
56 | 	def get_output(self):
57 | 		output_layer = self.layers[-1]
58 | 		res = [0] * output_layer.n_neurons
59 | 		for i in range(0, len(res)):
60 | 			res[i] = output_layer.output[i + use_bias]
61 | 		
62 | 		return res
63 | 		
64 | class Layer:
65 | 	
66 | 	def __init__(self, id, layer_size, prev_layer_size):
67 | 		self.id = id
68 | 		self.n_neurons = layer_size
69 | 		self.bias_val = 1
70 | 		
71 | 		self.input = [0] * self.n_neurons
72 | 		
73 | 		self.output = [0] * (self.n_neurons + 1)
74 | 		self.output[0] = self.bias_val
75 | 		
76 | 		self.error = [0] * self.n_neurons
77 | 		
78 | 		self.weight = make_matrix(prev_layer_size + use_bias, self.n_neurons)
79 | 		for i in range(len(self.weight)):
80 | 			for j in range(len(self.weight[i])):
81 | 				self.weight[i][j] = between(-0.2, 0.2)
82 | 		
83 | if __name__ == '__main__':
84 | 	# The AND function
85 | 	and_ann = ArtificialNeuralNetwork([2, 1])
86 | 	inputs = [[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]]
87 | 	targets = [[0.0], [0.0], [0.0], [1.0]]
88 | 	
89 | 	# The OR function
90 | 	or_ann = AritifialNeuralNetwork([2, 1])
91 | 	inputs = [[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]]
92 | 	targets = [[0.0], [1.0], [1.0], [1.0]]
93 | 	
94 | 	# Make predictions without training
95 | 	for i in range(len(targets)):
96 | 		print(and_ann.predict(inputs[i]))
97 | 


--------------------------------------------------------------------------------
/ArtificialNeuralNetworkUtility.py:
--------------------------------------------------------------------------------
 1 | import math
 2 | import random
 3 | 
 4 | def sigmoid(x):
 5 | 	"""
 6 | 	The sigmoid function. Returns a number between 0 and 1.
 7 | 	"""
 8 | 	return 1.0 / (1 + math.exp(-x))
 9 | 
10 | def between(min, max):
11 | 	"""
12 | 	Get a random value from the given interval
13 | 	"""
14 | 	return random.random() * (max - min) + min
15 | 
16 | def make_matrix(N, M):
17 | 	"""
18 | 	Make a matrix (N x M)
19 | 	"""
20 | 	return [[0 for i in range(M)] for i in range(N)]
21 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # Artificial Neural Networks (ANNs) in Python
2 | 
3 | Designing an artificial neural network from scratch.
4 | 


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