|
| 1 | +#include <iostream> |
| 2 | +#include <string> |
| 3 | +#include <math.h> |
| 4 | +#include "LogisticRegression.h" |
| 5 | +using namespace std; |
| 6 | + |
| 7 | + |
| 8 | +LogisticRegression::LogisticRegression(int size, int in, int out) { |
| 9 | + N = size; |
| 10 | + n_in = in; |
| 11 | + n_out = out; |
| 12 | + |
| 13 | + // initialize W, b |
| 14 | + W = new double*[n_in]; |
| 15 | + for (int i=0; i<n_in; i++) W[i] = new double[n_out]; |
| 16 | + b = new double[n_out]; |
| 17 | +} |
| 18 | + |
| 19 | +LogisticRegression::~LogisticRegression() { |
| 20 | + for (int i=0; i<n_out; i++) delete[] W[i]; |
| 21 | + delete[] W; |
| 22 | + delete[] b; |
| 23 | +} |
| 24 | + |
| 25 | + |
| 26 | +void LogisticRegression::train(int *x, int *y, double lr) { |
| 27 | + int i,j; |
| 28 | + double p_y_given_x[n_out]; |
| 29 | + double dy[n_out]; |
| 30 | + |
| 31 | + for (i=0; i<n_out; i++) { |
| 32 | + for (j=0; j<n_in; j++) { |
| 33 | + p_y_given_x[i] += W[i][j] * x[j]; |
| 34 | + } |
| 35 | + p_y_given_x[i] += b[i]; |
| 36 | + } |
| 37 | + softmax(p_y_given_x); |
| 38 | + |
| 39 | + for (i=0; i<n_out; i++) { |
| 40 | + dy[i] = y[i] - p_y_given_x[i]; |
| 41 | + |
| 42 | + for (j=0; j<n_in; j++) { |
| 43 | + W[i][j] += lr * dy[i] * x[j] / N; |
| 44 | + } |
| 45 | + |
| 46 | + b[i] += lr * dy[i] / N; |
| 47 | + } |
| 48 | +} |
| 49 | + |
| 50 | +void LogisticRegression::softmax(double *x) { |
| 51 | + double max; |
| 52 | + double sum; |
| 53 | + |
| 54 | + int i; |
| 55 | + for (i=0; i<n_out; i++) if(max < x[i]) max = x[i]; |
| 56 | + for (i=0; i<n_out; i++) { |
| 57 | + x[i] = exp(x[i] - max); |
| 58 | + sum += x[i]; |
| 59 | + } |
| 60 | + |
| 61 | + for(i=0; i<n_out; i++) x[i] /= sum; |
| 62 | +} |
| 63 | + |
| 64 | +void LogisticRegression::predict(int *x, double *y) { |
| 65 | + for (int i=0; i<n_out; i++) { |
| 66 | + for (int j=0; j<n_in; j++) { |
| 67 | + y[i] += W[i][j] * x[j]; |
| 68 | + } |
| 69 | + y[i] += b[i]; |
| 70 | + } |
| 71 | + |
| 72 | + softmax(y); |
| 73 | +} |
| 74 | + |
| 75 | + |
| 76 | +void test_lr() { |
| 77 | + int i,j; |
| 78 | + |
| 79 | + double learning_rate = 0.1; |
| 80 | + double n_epochs = 500; |
| 81 | + |
| 82 | + int train_N = 6; |
| 83 | + int test_N = 1; |
| 84 | + int n_in = 6; |
| 85 | + int n_out = 2; |
| 86 | + // int **train_X; |
| 87 | + // int **train_Y; |
| 88 | + // int **test_X; |
| 89 | + // double **test_Y; |
| 90 | + |
| 91 | + // train_X = new int*[train_N]; |
| 92 | + // train_Y = new int*[train_N]; |
| 93 | + // for (i=0; i<train_N; i++){ |
| 94 | + // train_X[i] = new int[n_in]; |
| 95 | + // train_Y[i] = new int[n_out]; |
| 96 | + // }; |
| 97 | + |
| 98 | + // test_X = new int*[test_N]; |
| 99 | + // test_Y = new double*[test_N]; |
| 100 | + // for (i=0; i<test_N; i++){ |
| 101 | + // test_X[i] = new int[n_in]; |
| 102 | + // test_Y[i] = new double[n_out]; |
| 103 | + // } |
| 104 | + |
| 105 | + |
| 106 | + // training data |
| 107 | + int train_X[6][6] = { |
| 108 | + {1, 1, 1, 0, 0, 0}, |
| 109 | + {1, 0, 1, 0, 0, 0}, |
| 110 | + {1, 1, 1, 0, 0, 0}, |
| 111 | + {0, 0, 1, 1, 1, 0}, |
| 112 | + {0, 0, 1, 1, 0, 0}, |
| 113 | + {0, 0, 1, 1, 1, 0} |
| 114 | + }; |
| 115 | + |
| 116 | + int train_Y[6][2] = { |
| 117 | + {1, 0}, |
| 118 | + {1, 0}, |
| 119 | + {1, 0}, |
| 120 | + {0, 1}, |
| 121 | + {0, 1}, |
| 122 | + {0, 1} |
| 123 | + }; |
| 124 | + |
| 125 | + |
| 126 | + // construct LogisticRegression |
| 127 | + LogisticRegression classifier(train_N, n_in, n_out); |
| 128 | + |
| 129 | + |
| 130 | + // train online |
| 131 | + for (int epoch=0; epoch<n_epochs; epoch++) { |
| 132 | + for (i=0; i<train_N; i++) { |
| 133 | + classifier.train(train_X[i], train_Y[i], learning_rate); |
| 134 | + } |
| 135 | + learning_rate *= 0.95; |
| 136 | + } |
| 137 | + |
| 138 | + |
| 139 | + // test data |
| 140 | + int test_X[1][6] = { |
| 141 | + {1, 1, 1, 0, 0, 0} |
| 142 | + }; |
| 143 | + |
| 144 | + double test_Y[1][2]; |
| 145 | + |
| 146 | + |
| 147 | + // test |
| 148 | + for (i=0; i<test_N; i++) { |
| 149 | + classifier.predict(test_X[i], test_Y[i]); |
| 150 | + for (j=0; j<n_out; j++) { |
| 151 | + cout << test_Y[i][j] << endl; |
| 152 | + } |
| 153 | + } |
| 154 | + |
| 155 | +} |
| 156 | + |
| 157 | + |
| 158 | +int main() { |
| 159 | + test_lr(); |
| 160 | + return 0; |
| 161 | +} |
0 commit comments