Predict with encoder + svm

spike_cluster_polisher.py

@@ -93,6 +93,7 @@ def filter_clusters (final_clusters, total_spikes, minimal_support=0.005):
 
   labels = [None] * total_spikes
   all_waves = None
+  encoded_waves = None
   index = 0
   thrown_spikes = 0
   kept_cluster = 0
@@ -107,8 +108,10 @@ def filter_clusters (final_clusters, total_spikes, minimal_support=0.005):
 
     if all_waves is None:
       all_waves = each[3]
+      encoded_waves = each[0]
     else:
       all_waves = np.vstack((all_waves, each[3]))
+      encoded_waves = np.vstack((encoded_waves, each[0]))
     for _ in range(len(each[0])):
       labels[index] = idx
       index += 1
@@ -118,4 +121,4 @@ def filter_clusters (final_clusters, total_spikes, minimal_support=0.005):
   if index + thrown_spikes != total_spikes:
     raise Exception ("Something is off %d %d %d" % (index, thrown_spikes, total_spikes))
 
-  return all_waves, labels, thrown_spikes
+  return encoded_waves, all_waves, labels, thrown_spikes

spike_main.py

@@ -222,19 +222,20 @@ def main ():
     logging.critical ("Done merging spikes. Total %d clusters!!!" % (len(final_clusters)))
 
     # Lastly, classify the results with SVM
-    all_waves, labels, n_thrown_spikes = filter_clusters(final_clusters, total_spikes)
+    encoded_waves, all_waves, labels, n_thrown_spikes = filter_clusters(final_clusters, total_spikes)
     logging.critical ("Done filtering. Ended up %d clusters and threw away %d spikes!!!" % (len(np.unique(labels)), n_thrown_spikes))
 
     logging.critical ("Started SVM classification!!!")
-    svm_classifier.Fit (data=all_waves, label=labels)
+    svm_classifier.Fit (data=encoded_waves, label=labels)
     logging.critical ("Done SVM classification!!!")
 
     scatter(all_waves, labels)
 
     # Consume trained SVMs
     new_spikes = get_sample_spikes ()
+    new_encoded_spikes = encoder.predict(np.array(new_spikes))
     new_labels = []
-    for each in new_spikes:
+    for each in new_encoded_spikes:
       res = svm_classifier.Predict (each)
       new_labels.append(res)