added test part

Author: Tomasz Latkowski

methods/selection.py

@@ -1,22 +1,6 @@
 import tensorflow as tf
 
 
-def selection_wrapper(data, num_instances, selection_method=None, num_features=None):
-    if data is None:
-        raise ValueError('Provide data to make selection.')
-
-    if selection_method is None:
-        raise ValueError('Provide selection method.')
-
-    if num_features is None:
-        data = tf.convert_to_tensor(data)
-        num_features = data.get_shape().as_list()[-1]
-
-    values, indices = selection_method(data, num_instances, num_features)
-    selected_features = tf.gather(data, indices, axis=1)
-    return values, selected_features
-
-
 def fisher(data, num_instances: list, top_k_features=2):
     """
     Performs Fisher feature selection method according to the following formula:

methods/selection_wrapper.py

@@ -14,5 +14,8 @@ def __init__(self, data, num_instances, selection_method=None, num_features=None
             data = tf.convert_to_tensor(data)
             num_features = data.get_shape().as_list()[-1]
 
-        self.values, indices = selection_method(data, num_instances, num_features)
-        self.selected_features = tf.gather(data, indices, axis=1)
+        self.values, self.indices = selection_method(data, num_instances, num_features)
+        self.selected_data = tf.gather(data, self.indices, axis=1)
+
+    def select(self, data):
+        return tf.gather(data, self.indices, axis=1)

run.py

@@ -13,6 +13,7 @@
 
 num_features = 100
 num_epochs = 1000
+eval_every = 10
 
 labels = np.concatenate([np.ones(82, dtype=np.float64), np.zeros(64, dtype=np.float64)])
 labels = np.reshape(labels, (-1, 1))
@@ -22,13 +23,16 @@
 
 for fold_id, (train_idxs, test_idxs) in enumerate(skf.split(data, labels.reshape(146))):
 
-    data_fold = data[train_idxs, :]
-    labels_fold = labels[train_idxs]
-    num_instances = [int(sum(labels_fold == 0)), int(sum(labels_fold == 1))]
+    data_train_fold = data[train_idxs, :]
+    labels_train_fold = labels[train_idxs]
+    num_instances = [int(sum(labels_train_fold == 0)), int(sum(labels_train_fold == 1))]
+
+    data_test_fold = data[test_idxs, :]
+    labels_test_fold = labels[test_idxs]
 
     with tf.Graph().as_default() as graph:
 
-        model = ExperimentModel(fisher, num_features, num_instances, None, data_fold)
+        model = ExperimentModel(fisher, num_features, num_instances, None, data_train_fold)
 
         with tf.Session() as session:
 
@@ -37,12 +41,21 @@
 
             log_saver = LogSaver('logs', 'fisher_fold{}'.format(fold_id), session.graph)
 
-            selected_data = session.run(model.selection_wrapper.selected_features)
+            train_selected_data = session.run(model.selection_wrapper.selected_data)
+            test_selected_data = session.run(model.selection_wrapper.select(data_test_fold))
 
             tqdm_iter = tqdm(range(num_epochs), desc='Epochs')
 
             for epoch in tqdm_iter:
-                feed_dict = {model.clf.x: selected_data, model.clf.y: labels_fold}
+                feed_dict = {model.clf.x: train_selected_data, model.clf.y: labels_train_fold}
                 loss, _, summary = session.run([model.clf.loss, model.clf.opt, model.clf.summary_op], feed_dict=feed_dict)
-                log_saver.log_train(summary, epoch)
+
+                if epoch % eval_every == 0:
+                    summary = session.run(model.clf.summary_op, feed_dict=feed_dict)
+                    log_saver.log_train(summary, epoch)
+
+                    feed_dict = {model.clf.x: test_selected_data, model.clf.y: labels_test_fold}
+                    summary = session.run(model.clf.summary_op, feed_dict=feed_dict)
+                    log_saver.log_test(summary, epoch)
+
                 tqdm_iter.set_postfix(loss='{:.2f}'.format(float(loss)), epoch=epoch)