-
Notifications
You must be signed in to change notification settings - Fork 0
/
CHL.py
81 lines (67 loc) · 2.83 KB
/
CHL.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
import numpy as np
def sigmoid(x):
return 1 / (1 + np.exp(-x))
# Neural Network with contrastive hebbian learning
class CHLNeuralNetwork:
# Constructor Function
# layerSizes: List of integers describing the size of each sequential layer
def __init__(self, layerSizes):
self.layerSizes = layerSizes
self.L = len(layerSizes)-1
self.B = [np.zeros((x,1)) for x in layerSizes]
weightShapes = [(a,b) for a,b in zip(layerSizes[:-1],layerSizes[1:])]
self.W = [np.zeros((1,1))] + [np.random.standard_normal(s) / s[1] ** .5 for s in weightShapes]
# Prediction Function
# a: Input data to feed through layers
def predict(self, x):
for k in range(1,self.L+1):
x = sigmoid(np.matmul(self.W[k].T,x) + self.B[k])
return x
# Printing helper function
def printAccuracy(self, trainingData, testingData):
correct = 0
for x,y in trainingData:
prediction = self.predict(x)
if np.argmax(prediction) == np.argmax(y[:,0]):
correct += 1
print('Training Acc: {0}/{1} ({2}%)'.format(correct, len(trainingData), (correct / len(trainingData)) * 100))
correct = 0
for x,y in testingData:
prediction = self.predict(x)
if np.argmax(prediction) == np.argmax(y[:,0]):
correct += 1
print('Testing Acc: {0}/{1} ({2}%)'.format(correct, len(testingData), (correct / len(testingData)) * 100))
# Contrastive Hebbian Learning algorithm
def chl(self, x, y, learningRate, feedback):
# Free phase
Xf = [np.zeros((x,1)) for x in self.layerSizes]
Xf[0] = x
for _ in range(50):
for k in range(1,self.L):
Xf[k] = sigmoid(np.matmul(self.W[k].T,Xf[k-1]) + (feedback * np.matmul(self.W[k+1],Xf[k+1])) + self.B[k])
Xf[self.L] = sigmoid(np.matmul(self.W[self.L].T,Xf[self.L-1]) + self.B[self.L])
# Clamped phase
Xc = [np.zeros((x,1)) for x in self.layerSizes]
Xc[0] = x
Xc[self.L] = y
for _ in range(50):
for k in range(1, self.L):
Xc[k] = sigmoid(np.matmul(self.W[k].T,Xc[k-1]) + (feedback * np.matmul(self.W[k+1],Xc[k+1])) + self.B[k])
# Update Weights and Biases
for k in range(1,self.L+1):
self.W[k] += learningRate * (feedback**(k-self.L)) * (np.matmul(Xc[k],Xc[k-1].T) - np.matmul(Xf[k],Xf[k-1].T)).T
self.B[k] += learningRate * (feedback**(k-self.L)) * (Xc[k] - Xf[k])
# Train NN via contrastive hebbian learning
def train(self, trainingData, testingData, epochs, learningRate):
for j in range(epochs):
# Shuffle training data to provide different order for each epoch
# random.shuffle(trainingData)
for x,y in trainingData:
self.chl(x, y, learningRate, 0.5)
# Test current accuracy
print("Epoch {0} complete".format(j+1))
self.printAccuracy(trainingData, testingData)
# Modify for debugging purposes
def test(self, x,y):
self.chl(x,y,1,1)
self.chl(x,y,1,1)