MLib/FeedForwardLayer.cpp at main · William-Baker/MLib · GitHub

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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
#include "FeedForwardLayer.hpp"
#include "ConvolutionalLayer.hpp"


void FeedForwardLayer::randomise() {
    weights.randomFill(-0.3, -0.05, 0.05, 0.3);
    biases.randomFill(-0.3, -0.05, 0.05, 0.3);
    //weights.randomFill(0, 1);
    //biases.randomFill(0, 1);
}

void FeedForwardLayer::zeros(){
    weights.fill(0);
    weights_deltas.fill(0);
    biases.fill(0);
    output.fill(0);
    error_signal.fill(0);
    net.fill(0);
    temp2.fill(0);
    gradient.fill(0);
}

// void compute() {
// 	Matrix tmp(get_prev_layer()->get_output(), false);
// 	compute(tmp);
// }

void FeedForwardLayer::compute(MLStruct<double>* inputIn) {
    delete input;
    input = new Matrix(inputIn, false);
    compute();
}
void FeedForwardLayer::compute() {
    weights.multiply(*input, output);
    output.add(biases, net);
    net.sigmoid(output);
    if (get_next_layer() != 0) {
        get_next_layer()->compute();
    }
}

void FeedForwardLayer::calculateErrorSignal(Matrix& outputError) {
    weights.multiplyA(outputError, error_signal);
    /*std::cout << "------- Output Error ------------" << std::endl;
    outputError.print();
    std::cout << "------- Weights ------------" << std::endl;
    weights.print();
    std::cout << "------- Error Signal ------------" << std::endl;
    errorSig.print();
    std::cout << std::endl;
    std::cout << std::endl;
    */
}

void FeedForwardLayer::update(double LR, Matrix& errorSigAtOutput) {
    output.sigmoidDifferential(gradient);
    gradient.scale(LR, temp2);
    temp2.multiplyElementWise(errorSigAtOutput, gradient);
    gradient.multiplyB(*input, weights_deltas);
    weights.addAssign(weights_deltas);
    //std::cout << "------- Weights Deltas ------------" << std::endl;
    //weights_deltas.print();
    biases.addAssign(gradient);
}

void FeedForwardLayer::backprop(double LR) {
    calculateErrorSignal(*error_signal_next_layer);
    update(LR, *error_signal_next_layer);
    get_prev_layer()->backprop(LR);
}
/**
 * @deprecated
 */
void FeedForwardLayer::backprop(MLStruct<double>* err, double LR) {
    ilog(WARNING, "backprop call deprecated, should connect to next layer instead");
    Matrix tmp(err,false);
    backprop(tmp, LR);
}
/**
 * @deprecated
 */
void FeedForwardLayer::backprop(Matrix& outErrorSig, double LR)  {
    ilog(WARNING, "backprop call deprecated, should connect to next layer instead");
    calculateErrorSignal(outErrorSig);
    update(LR, outErrorSig);
    get_prev_layer()->backprop(LR);
}


void FeedForwardLayer::print() {

    std::cout << "--------------- FF Weights Layer Y: " << weights.height() << " X: " << weights.width() <<  " ---------------" << std::endl;
    weights.print();
    std::cout << "--------------- FF Weights Deltas Layer Y: " << weights_deltas.height() << " X: " << weights_deltas.width() << " ---------------" << std::endl;
    weights_deltas.print();
    std::cout << "--------------- FF Biases Layer Y: " << biases.height() << " ---------------" << std::endl;
    biases.print();
    std::cout << "--------------- FF Output Layer Y: " << output.height() << " ---------------" << std::endl;
    output.print();
    std::cout << "--------------- FF Out Error Layer Y: " << error_signal.height() << " ---------------" << std::endl;
    error_signal.print();

}


void FeedForwardLayer::connect_to_next_layer(Layer* next){
    if(dynamic_cast<OutputLayer*>(next)){
        OutputLayer* layer = static_cast<OutputLayer*>(next);
        error_signal_next_layer = new Matrix(layer->get_error_signal(), false);
        error_signal_next_layer->resize(biases.size(),1);
    }
    else if(dynamic_cast<FeedForwardLayer*>(next)){
        FeedForwardLayer* layer = static_cast<FeedForwardLayer*>(next);
        error_signal_next_layer = new Matrix(layer->get_error_signal(), false);
        error_signal_next_layer->resize(biases.size(),1);
    }
    else if(dynamic_cast<ConvolutionalLayer*>(next)){
        ConvolutionalLayer* layer = static_cast<ConvolutionalLayer*>(next);
        error_signal_next_layer = new Matrix(Matrix(layer->get_error_signal(), false).copy_keeping_same_data());
        error_signal_next_layer->resize(biases.size(),1);
    }
    else{
        ilog(FATAL_ERROR, "unrecognized next layer");
    }
}

void FeedForwardLayer::connect_to_prev_layer(Layer* prev){
    if(dynamic_cast<InputLayer*>(prev)){
        InputLayer* layer = static_cast<InputLayer*>(prev);
    }
    else if(dynamic_cast<FeedForwardLayer*>(prev)){
        FeedForwardLayer* layer = static_cast<FeedForwardLayer*>(prev);
        input = new Matrix(Matrix(layer->get_output(), false).copy_keeping_same_data());
        input->resize(this->error_signal.size(), 1); //use error signal to resize

    }
    else if(dynamic_cast<ConvolutionalLayer*>(prev)){
        ConvolutionalLayer* layer = static_cast<ConvolutionalLayer*>(prev);
        input = new Matrix(Matrix(layer->get_output(), false).copy_keeping_same_data());
        input->resize(weights.width(),1);
    }
    else{
        ilog(FATAL_ERROR, "unrecognized previous layer");
    }
}