Add JS template to support exported model data

sklearn_porter/estimator/classifier/KNeighborsClassifier/templates/js/exported.class.txt

@@ -0,0 +1,115 @@
+if (typeof XMLHttpRequest === 'undefined') {{
+ var XMLHttpRequest = require("xmlhttprequest").XMLHttpRequest;
+}}
+
+var {class_name} = function(jsonFile) {{
+ this.data = undefined;
+
+ var Neighbor = function(y, dist) {{
+ this.y = y;
+ this.dist = dist;
+ }};
+
+ var promise = new Promise(function(resolve, reject) {{
+ var httpRequest = new XMLHttpRequest();
+ httpRequest.onreadystatechange = function() {{
+ if (httpRequest.readyState === 4) {{
+ if (httpRequest.status === 200) {{
+ resolve(JSON.parse(httpRequest.responseText));
+ }} else {{
+ reject(new Error(httpRequest.statusText));
+ }}
+ }}
+ }};
+ httpRequest.open('GET', jsonFile, true);
+ httpRequest.send();
+ }});
+
+ var compute = function(temp, cand, q) {{
+ var dist = 0.,
+ diff;
+ for (var i = 0, l = temp.length; i < l; i++) {{
+ diff = Math.abs(temp[i] - cand[i]);
+ if (q==1) {{
+ dist += diff;
+ }} else if (q==2) {{
+ dist += diff*diff;
+ }} else if (q==Number.POSITIVE_INFINITY) {{
+ if (diff > dist) {{
+ dist = diff;
+ }}
+ }} else {{
+ dist += Math.pow(diff, q);
+ }}
+ }}
+ if (q==1 || q==Number.POSITIVE_INFINITY) {{
+ return dist;
+ }} else if (q==2) {{
+ return Math.sqrt(dist);
+ }} else {{
+ return Math.pow(dist, 1. / q);
+ }}
+ }};
+
+ this.{method_name} = function(features) {{
+ return new Promise(function(resolve, reject) {{
+ promise.then(function(data) {{
+ if (typeof this.data === 'undefined') {{
+ this.data = data;
+ this.nTemplates = this.data.X.length;
+ }}
+ var classIdx = 0, i, dist;
+ if (this.data.nNeighbors == 1) {{
+ var minDist = Number.POSITIVE_INFINITY;
+ for (i = 0; i < this.data.nTemplates; i++) {{
+ dist = compute(this.data.X[i], features, this.data.power);
+ if (dist <= minDist) {{
+ minDist = dist;
+ classIdx = this.data.y[i];
+ }}
+ }}
+ }} else {{
+ var classes = new Array(this.data.nClasses).fill(0);
+ var dists = [];
+ for (i = 0; i < this.nTemplates; i++) {{
+ dist = compute(this.data.X[i], features, this.data.power);
+ dists.push(new Neighbor(this.data.y[i], dist));
+ }}
+ dists.sort(function compare(n1, n2) {{
+ return (n1.dist < n2.dist) ? -1 : 1;
+ }});
+ for (i = 0; i < this.data.kNeighbors; i++) {{
+ classes[dists[i].y]++;
+ }}
+ for (i = 0; i < this.data.nClasses; i++) {{
+ classIdx = classes[i] > classes[classIdx] ? i : classIdx;
+ }}
+ }}
+ resolve(classIdx);
+ }}, function(error) {{
+ reject(error);
+ }});
+ }});
+ }};
+
+}};
+
+if (typeof process !== 'undefined' && typeof process.argv !== 'undefined') {{
+ if (process.argv[2].trim().endsWith('.json')) {{
+
+ // Parameters:
+ var json = process.argv[2];
+
+ // Estimator:
+ var clf = new {class_name}(json);
+
+ // Features:
+ var features = process.argv.slice(3);
+
+ // Prediction:
+ clf.{method_name}(features).then(function(prediction) {{
+ console.log(prediction);
+ }});
+
+ }}
+}}

tests/estimator/classifier/KNeighborsClassifier/KNeighborsClassifierJSTest.py

@@ -2,6 +2,7 @@
 
 from unittest import TestCase
 
+import numpy as np
 from sklearn.neighbors import KNeighborsClassifier
 
 from tests.estimator.classifier.Classifier import Classifier
@@ -16,3 +17,81 @@ def setUp(self):
 
  def tearDown(self):
  super(KNeighborsClassifierJSTest, self).tearDown()
+
+ def test_random_features_w_binary_data_export(self):
+ self.load_binary_data()
+ self._port_estimator(export_data=True)
+ amin = np.amin(self.X, axis=0)
+ amax = np.amax(self.X, axis=0)
+ preds, ground_truth = [], []
+ for _ in range(self.N_RANDOM_FEATURE_SETS):
+ x = np.random.uniform(amin, amax, self.n_features)
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)
+
+ def test_random_features_w_iris_data_export(self):
+ self.load_iris_data()
+ self._port_estimator(export_data=True)
+ amin = np.amin(self.X, axis=0)
+ amax = np.amax(self.X, axis=0)
+ preds, ground_truth = [], []
+ for _ in range(self.N_RANDOM_FEATURE_SETS):
+ x = np.random.uniform(amin, amax, self.n_features)
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)
+
+ def test_random_features_w_digits_data_export(self):
+ self.load_digits_data()
+ self._port_estimator(export_data=True)
+ amin = np.amin(self.X, axis=0)
+ amax = np.amax(self.X, axis=0)
+ preds, ground_truth = [], []
+ for _ in range(self.N_RANDOM_FEATURE_SETS):
+ x = np.random.uniform(amin, amax, self.n_features)
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)
+
+ def test_existing_features_w_binary_data_export(self):
+ self.load_binary_data()
+ self._port_estimator(export_data=True)
+ preds, ground_truth = [], []
+ n = min(self.N_EXISTING_FEATURE_SETS, len(self.X))
+ for x in self.X[:n]:
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)
+
+ def test_existing_features_w_iris_data_export(self):
+ self.load_iris_data()
+ self._port_estimator(export_data=True)
+ preds, ground_truth = [], []
+ n = min(self.N_EXISTING_FEATURE_SETS, len(self.X))
+ for x in self.X[:n]:
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)
+
+ def test_existing_features_w_digits_data_export(self):
+ self.load_digits_data()
+ self._port_estimator(export_data=True)
+ preds, ground_truth = [], []
+ n = min(self.N_EXISTING_FEATURE_SETS, len(self.X))
+ for x in self.X[:n]:
+ preds.append(self.pred_in_custom(x, export_data=True))
+ ground_truth.append(self.pred_in_py(x))
+ self._clear_estimator()
+ # noinspection PyUnresolvedReferences
+ self.assertListEqual(preds, ground_truth)