basic implementation of probability estimates

#9

Makefile

@@ -3,7 +3,7 @@ CXX = em++
 
 CFLAGS = -Wall -Wconversion -O3 -fPIC --memory-init-file 0
 BUILD_DIR=dist
-EXPORTED_FUNCTIONS="['_parse_command_line', '_create_svm_nodes', '_add_instance', '_libsvm_train_problem', '_libsvm_train', '_libsvm_predict_one', '_get_svr_epsilon', '_svm_free_model', '_svm_get_svm_type', '_svm_get_nr_sv', '_svm_get_nr_class', '_svm_get_sv_indices', '_svm_get_labels', '_libsvm_cross_validation', '_free_problem', '_serialize_model', '_deserialize_model']"
+EXPORTED_FUNCTIONS="['_parse_command_line', '_create_svm_nodes', '_add_instance', '_libsvm_train_problem', '_libsvm_train', '_libsvm_predict_one', '_libsvm_predict_one_probability', '_get_svr_epsilon', '_svm_free_model', '_svm_get_svm_type', '_svm_get_nr_sv', '_svm_get_nr_class', '_svm_get_sv_indices', '_svm_get_labels', '_libsvm_cross_validation', '_free_problem', '_serialize_model', '_deserialize_model']"
 
 all: wasm asm
 

examples/probabilities.js

@@ -0,0 +1,41 @@
+const Kernel = new require('ml-kernel');
+const SVM = require('../asm');
+const range = require('lodash.range');
+
+'use strict';
+
+const gamma = 0.2;
+const cost = 1;
+
+function exec(SVM, precomputed) {
+    const data = require('ml-dataset-iris');
+    var trainData;
+
+    const features = data.getNumbers();
+    let labels = data.getClasses();
+    const classes = data.getDistinctClasses();
+    const c = {};
+    classes.forEach((v, idx) => c[v] = idx);
+    labels = labels.map(l => c[l]);
+
+
+    if (precomputed) {
+        const kernel = new Kernel('gaussian', {sigma: 1 / Math.sqrt(gamma)});
+        trainData = kernel.compute(features).addColumn(0, range(1, labels.length + 1));
+    } else {
+        trainData = features;
+    }
+
+    const svm = new SVM({
+        quiet: true,
+        cost: cost,
+        kernel: precomputed ? SVM.KERNEL_TYPES.PRECOMPUTED : SVM.KERNEL_TYPES.RBF,
+        gamma,
+        probabilityEstimates: true
+    });
+    svm.train(trainData, labels);
+    var pred = svm.predictProbability(trainData);
+    console.log(JSON.stringify(pred, null, 2));
+}
+
+exec(SVM, true);

js-interfaces.c

@@ -194,18 +194,29 @@ void free_problem(struct svm_problem* prob) {
     free(prob);
 }
 
-double libsvm_predict_one(struct svm_model* model, double* data, int size) {
+struct svm_node* init_node(double* data, int size) {
     struct svm_node* node = Malloc(struct svm_node, size + 1);
     for(int i=0; i<size; i++) {
         node[i].index = i + 1;
         node[i].value = data[i];
     }
     node[size].index = -1;
+    return node;
+}
+
+double libsvm_predict_one(struct svm_model* model, double* data, int size) {
+    struct svm_node* node = init_node(data, size);
     double pred = svm_predict(model, node);
     free(node);
     return pred;
 }
 
+double libsvm_predict_one_probability(struct svm_model* model, double* data, int size, double* prob_estimates) {
+    struct svm_node* node = init_node(data, size);
+    double pred = svm_predict_probability(model, node, prob_estimates);
+    return pred;
+}
+
 struct svm_model* libsvm_train(double *data, double *labels, int nb_features, int nb_dimensions, const char* command) {
     struct svm_problem* prob = create_svm_nodes(nb_features, nb_dimensions);
     for(int i=0; i<nb_features; i++) {

src/loadSVM.js

@@ -3,6 +3,7 @@ module.exports = function (libsvm) {
 
     /* eslint-disable camelcase */
     const predict_one = libsvm.cwrap('libsvm_predict_one', 'number', ['number', 'array', 'number']);
+    const predict_one_probability = libsvm.cwrap('libsvm_predict_one_probability', 'number', ['number', 'array', 'number', 'number']);
     const add_instance = libsvm.cwrap('add_instance', null, ['number', 'array', 'number', 'number', 'number']);
     const create_svm_nodes = libsvm.cwrap('create_svm_nodes', 'number', ['number', 'number']);
     const train_problem = libsvm.cwrap('libsvm_train_problem', 'number', ['number', 'string']);
@@ -131,6 +132,42 @@ module.exports = function (libsvm) {
             return arr;
         }
 
+        /**
+         * Predict the label with probability estimate of many samples.
+         * @param {Array<Array<number>>} samples - The samples to predict.
+         * @return {Array<object>} - An array of objects containing the prediction label and the probability estimates for each label
+         */
+        predictProbability(samples) {
+            let arr = [];
+            for (let i = 0; i < samples.length; i++) {
+                arr.push(this.predictOneProbability(samples[i]));
+            }
+            return arr;
+        }
+
+        /** Predict the label with probability estimate.
+         * @param {Array<number>} sample
+         * @return {object} - An object containing the prediction label and the probability estimates for each label
+         */
+
+        predictOneProbability(sample) {
+            const labels = this.getLabels();
+            const nbLabels = labels.length;
+            const estimates = libsvm._malloc(nbLabels * 8);
+            const prediction = predict_one_probability(this.model, new Uint8Array(new Float64Array(sample).buffer), sample.length, estimates);
+            const estimatesArr = Array.from(libsvm.HEAPF64.subarray(estimates / 8, estimates / 8 + nbLabels));
+            const result = {
+                prediction,
+                estimates: labels.map((label, idx) => ({
+                    label,
+                    probability: estimatesArr[idx]
+                }))
+            };
+            libsvm._free(estimates);
+            return result;
+        }
+
+
         /**
          * Get the array of labels from the model. Useful when creating an SVM instance with SVM.load
          * @return {Array<number>} - The list of labels.