Line data Source code
1 : /**
2 : * @file gensvm_predict.c
3 : * @author G.J.J. van den Burg
4 : * @date 2013-08-09
5 : * @brief Main functions for predicting class labels..
6 : *
7 : * @details
8 : * This file contains functions for predicting the class labels of instances
9 : * and a function for calculating the predictive performance (hitrate) of
10 : * a prediction given true class labels.
11 : *
12 : * @copyright
13 : Copyright 2016, G.J.J. van den Burg.
14 :
15 : This file is part of GenSVM.
16 :
17 : GenSVM is free software: you can redistribute it and/or modify
18 : it under the terms of the GNU General Public License as published by
19 : the Free Software Foundation, either version 3 of the License, or
20 : (at your option) any later version.
21 :
22 : GenSVM is distributed in the hope that it will be useful,
23 : but WITHOUT ANY WARRANTY; without even the implied warranty of
24 : MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25 : GNU General Public License for more details.
26 :
27 : You should have received a copy of the GNU General Public License
28 : along with GenSVM. If not, see <http://www.gnu.org/licenses/>.
29 :
30 : */
31 :
32 : #include "gensvm_predict.h"
33 :
34 : /**
35 : * @brief Predict class labels of data given and output in predy
36 : *
37 : * @details
38 : * The labels are predicted by mapping each instance in data to the
39 : * simplex space using the matrix V in the given model. Next, for each
40 : * instance the nearest simplex vertex is determined using an Euclidean
41 : * norm. The nearest simplex vertex determines the predicted class label,
42 : * which is recorded in predy.
43 : *
44 : * @param[in] testdata GenData to predict labels for
45 : * @param[in] model GenModel with optimized V
46 : * @param[out] predy pre-allocated vector to record predictions in
47 : */
48 2 : void gensvm_predict_labels(struct GenData *testdata, struct GenModel *model,
49 : long *predy)
50 : {
51 : long i, j, k, n, K, label;
52 : double norm, min_dist,
53 2 : *S = NULL,
54 2 : *ZV = NULL;
55 :
56 2 : n = testdata->n;
57 2 : K = model->K;
58 :
59 : // allocate necessary memory
60 2 : S = Calloc(double, K-1);
61 2 : ZV = Calloc(double, n*(K-1));
62 :
63 : // Generate the simplex matrix
64 2 : gensvm_simplex(model);
65 :
66 : // Generate the simplex space vectors
67 2 : gensvm_calculate_ZV(model, testdata, ZV);
68 :
69 : // Calculate the distance to each of the vertices of the simplex.
70 : // The closest vertex defines the class label
71 26 : for (i=0; i<n; i++) {
72 24 : label = 0;
73 24 : min_dist = INFINITY;
74 96 : for (j=0; j<K; j++) {
75 216 : for (k=0; k<K-1; k++) {
76 288 : S[k] = matrix_get(ZV, K-1, i, k) -
77 144 : matrix_get(model->U, K-1, j, k);
78 : }
79 72 : norm = cblas_dnrm2(K-1, S, 1);
80 72 : if (norm < min_dist) {
81 44 : label = j+1;
82 44 : min_dist = norm;
83 : }
84 : }
85 24 : predy[i] = label;
86 : }
87 :
88 2 : free(ZV);
89 2 : free(S);
90 2 : }
91 :
92 : /**
93 : * @brief Calculate the predictive performance (percentage correct)
94 : *
95 : * @details
96 : * The predictive performance is calculated by simply counting the number
97 : * of correctly classified samples and dividing by the total number of
98 : * samples, multiplying by 100.
99 : *
100 : * @param[in] data the GenData dataset with known labels
101 : * @param[in] predy the predicted class labels
102 : *
103 : * @returns percentage correctly classified.
104 : */
105 3 : double gensvm_prediction_perf(struct GenData *data, long *predy)
106 : {
107 3 : long i, correct = 0;
108 : double performance;
109 :
110 27 : for (i=0; i<data->n; i++)
111 24 : if (data->y[i] == predy[i])
112 8 : correct++;
113 :
114 3 : performance = ((double) correct)/((double) data->n)* 100.0;
115 :
116 3 : return performance;
117 : }
|
