File size: 3,248 Bytes
d4c7aae | 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 | import numpy as np
# https://medium.com/towards-data-science/math-neural-network-from-scratch-in-python-d6da9f29ce65
class Network:
def __init__(self):
self.layers = []
self.loss = None
self.loss_prime = None
self.weight_list = []
def add(self, layer):
self.layers.append(layer)
def use(self, loss, loss_prime):
self.loss = loss
self.loss_prime = loss_prime
def fit(self, x_train, y_train, epochs, learning_rate):
samples = len(x_train)
# Train epoch times
for i in range(epochs):
err = 0
# train for all samples
for j in range(samples):
input = x_train[j]
output = None
# predict data at all layer
for layer in self.layers:
output = layer.forward_propagation(input)
# output as input of next layer
input = output
# Find loss for display
err += self.loss(y_train[j], output)
# Find Error of output dE/dY using derivation of MSE
error = self.loss_prime(y_train[j], output)
# Update Weight for next data
# By find gradient of weight in each layer and update
for layer in reversed(self.layers):
error = layer.backward_propagation(error, learning_rate)
# Find Average Error per sample
err /= samples
print("Epoch %d/%d calculate with error = %f" %
(i + 1, epochs, err))
print(f"Update weight to {layer.get_weight()} ")
print(f"Update Bias to {layer.get_bias()}")
print("")
def fit_on_sample(self):
raise NotImplementedError
def get_weight_list(self):
return self.weight_list
def get_weights(self):
raise NotImplementedError
def get_biases(self):
result = []
for layer in self.layers:
result.append(layer.get_bias())
return result
def predict_sample(self, first_input, second_input):
output = None
# predict data at all layer
for layer in self.layers:
output = layer.forward_propagation(first_input, second_input)
# output as input of next layer
first_input = output
return output
def back_propagate(self, error, learning_rate):
# weight = []
for layer in reversed(self.layers):
error = layer.backward_propagation(error, learning_rate)
# weight.append(layer.get_weight())
# print(f'Epoch calculate with error = {err}')
# self.weight_list = np.append(self.weight_list, [weight])
def predict(self, first_input, second_input):
output = None
# predict data at all layer
for layer in self.layers:
output = layer.predict(first_input, second_input)
# output as input of next layer
first_input = output
return output
def check_type(self):
print(
"It is in the Initial Class Network Please call this function on the child class"
)
return False
|