| 79 | // ---------------------------------------------------------------------------------------- |
| 80 | |
| 81 | int predict_label ( |
| 82 | const column_vector& weights, |
| 83 | const sample_type& sample |
| 84 | ) |
| 85 | /*! |
| 86 | requires |
| 87 | - weights.size() == 9 |
| 88 | - sample.size() == 3 |
| 89 | ensures |
| 90 | - Given the 9-dimensional weight vector which defines a 3 class classifier, this |
| 91 | function predicts the class of the given 3-dimensional sample vector. |
| 92 | Therefore, the output of this function is either 0, 1, or 2 (i.e. one of the |
| 93 | three possible labels). |
| 94 | !*/ |
| 95 | { |
| 96 | // Our 3-class classifier model can be thought of as containing 3 separate linear |
| 97 | // classifiers. So to predict the class of a sample vector we evaluate each of these |
| 98 | // three classifiers and then whatever classifier has the largest output "wins" and |
| 99 | // predicts the label of the sample. This is the popular one-vs-all multi-class |
| 100 | // classifier model. |
| 101 | // |
| 102 | // Keeping this in mind, the code below simply pulls the three separate weight vectors |
| 103 | // out of weights and then evaluates each against sample. The individual classifier |
| 104 | // scores are stored in scores and the highest scoring index is returned as the label. |
| 105 | column_vector w0, w1, w2; |
| 106 | w0 = rowm(weights, range(0,2)); |
| 107 | w1 = rowm(weights, range(3,5)); |
| 108 | w2 = rowm(weights, range(6,8)); |
| 109 | |
| 110 | column_vector scores(3); |
| 111 | scores = dot(w0, sample), dot(w1, sample), dot(w2, sample); |
| 112 | |
| 113 | return index_of_max(scores); |
| 114 | } |
| 115 | |
| 116 | // ---------------------------------------------------------------------------------------- |
| 117 | // ---------------------------------------------------------------------------------------- |
no test coverage detected