-
Notifications
You must be signed in to change notification settings - Fork 1
/
neural_network_6_Sai.py
110 lines (92 loc) · 3.94 KB
/
neural_network_6_Sai.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
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Feb 18 19:29:33 2021
@author: saidivyesh, varun
"""
from PIL import Image
import numpy as np
import math
class NeuralNetwork:
# L = # of layers
# sizes = []*L (number of neurons in each layer)
# network = []*L (network[i] is an array with all of the neuron
# values)
# weights = []*(L-1) (weights[i] is a matrix with all of the
# weights between each pair of neurons
# in each layer)
# biases = []*(L-1) (biases[i] is an array with all of the
# biases of the (i+1)th layer)
# error = []*L (basically network but with all of the errors)
# delta = []*L (basically error but with the derivatives of the errors)
def __init__(self, L, sizes, inp, w_1, w_2, b_1, b_2):
self.L = L
self.sizes = sizes
self.network = []*L
self.error = []*L
self.delta = []*L
self.weights = [[], []]
self.biases = [[], []]
self.network = [[]*sizes[0], []*sizes[1], []*sizes[2]]
#for i in range(0, L-1):
# self.network[i] = []*sizes[i]
self.weights[0] = w_1
self.weights[1] = w_2
self.biases[0] = b_1
self.biases[1] = b_2
#self.network[L-1] = []*sizes[L-1]
self.network[0] = inp
def forward(self):
for l in range(1, self.L):
for i in range(0, self.sizes[l]):
self.network[l][i] = self.biases[l-1][i];
for j in range(0, self.sizes[l-1]):
self.network[l][i] += self.weights[l-1][j][i] * self.network[l-1][j]
# for i in range(0, self.sizes[l]):
# self.network[l][i] = (1/(1 + np.exp(-self.network[l][i])))
return self.network[self.L - 1]
def derivative(output):
return (output) * (1 - output)
def backward(self, expected_out):
for l in reversed(range(1, self.L)):
if (l == self.L - 1):
for i in range(0, self.sizes[l]):
self.error[l][i] = (self.expected_out[i] - self.network[l][i])
else:
for i in range(0, self.sizes[l]):
for j in range (0, self.sizes[l+1]):
self.error[l][i] += self.weights[l][i][j] * self.delta[l+1][j]
self.error[l][i] += self.biases[l-1][i]
for i in range(0, self.sizes[l]):
self.delta[l][i] = self.error[l][i] * self.derivative(self.network[l][i])
# 0.25 = learning rate (we will run in batches)
def update_weights(self, learning_rate):
for l in range(1, self.L):
for i in range(0, self.sizes[l]):
for j in range(self.sizes[l-1]):
self.weights[l-1][j][i] += learning_rate * self.network[l-1][j] * self.delta[l][i]
self.biases[l-1][i] += learning_rate * self.delta[l][i]
# collect all the data, split into train and test
#
def train_network(self, train, learning_rate, epochs, expected):
for epoch in range(0, epochs):
cost_sum = 0
for inp in train:
out = self.forward()
expected_out = [0 for i in range(expected)]
# expected_out[inp[-1]] = 1
cost_sum += sum([(expected_out[i]-out[i])**2 for i in range(len(expected))])
self.backward(expected_out)
self.update_weights(learning_rate)
n_1 = 412368
n_2 = 10000
n_3 = 10
image = Image.open("/Users/saimonish/IntelliJ_workspace/ACSEF2021/CroppedImages/dutmc_09_1_cropped.png")
input_values = list(image.getdata())
w_1 = [[0.5]*10000]*412368
w_2 = [[1, 1, 1], [1, 1, 1], [1, 1, 1], [1, 1, 1]]
b_1 = [0]*10000
b_2 = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
#nn = NeuralNetwork(n_1, n_2, n_3, inp, w_1, w_2, b_1, b_2)
nn = NeuralNetwork(3, [n_1, n_2, n_3], input_values, w_1, w_2, b_1, b_2)
nn.train_network([input_values], 0.1, 1, 1)