Classifier: infer() returns empty array if unlearned

instead of assertion, that makes processing pipelines easier.

bindings/py/tests/algorithms/sdr_classifier_test.py

@@ -18,6 +18,7 @@
 
 """ Unit tests for Classifier & Predictor classes. """
 
+import math
 import numpy
 import pickle
 import random
@@ -131,14 +132,13 @@ def testComputeInferOrLearnOnly(self):
     inp.randomize( .3 )
 
     # learn only
-    with self.assertRaises(RuntimeError):
-      c.infer(pattern=inp) # crash with not enough training data.
+    prediction = c.infer(pattern=inp)[1]
+    self.assertTrue(prediction == []) # not enough training data -> []
     c.learn(recordNum=0, pattern=inp, classification=4)
-    with self.assertRaises(RuntimeError):
-      c.infer(pattern=inp) # crash with not enough training data.
+    self.assertTrue(c.infer(pattern=inp)[1] == []) # not enough training data.
     c.learn(recordNum=2, pattern=inp, classification=4)
     c.learn(recordNum=3, pattern=inp, classification=4)
-    c.infer(pattern=inp) # Don't crash with not enough training data.
+    self.assertTrue(c.infer(pattern=inp)[1] != []) # Don't crash with enough training data.
 
     # infer only
     retval1 = c.infer(pattern=inp)

py/htm/examples/hotgym.py

@@ -158,17 +158,16 @@ def main(parameters=default_parameters, argv=None, verbose=True):
     tm_info.addData( tm.getActiveCells().flatten() )
 
     # Predict what will happen, and then train the predictor based on what just happened.
-    if count > 5: #skip the n(=to the furthest predictions step) step, as predictor must learn something first
-      pdf = predictor.infer( tm.getActiveCells() )
-      for n in (1, 5):
-        if pdf[n]:
-          predictions[n].append( np.argmax( pdf[n] ) * predictor_resolution )
-        else:
-          predictions[n].append(float('nan'))
-
-      anomalyLikelihood = anomaly_history.anomalyProbability( consumption, tm.anomaly )
-      anomaly.append( tm.anomaly )
-      anomalyProb.append( anomalyLikelihood )
+    pdf = predictor.infer( tm.getActiveCells() )
+    for n in (1, 5):
+      if pdf[n]:
+        predictions[n].append( np.argmax( pdf[n] ) * predictor_resolution )
+      else:
+        predictions[n].append(float('nan'))
+
+    anomalyLikelihood = anomaly_history.anomalyProbability( consumption, tm.anomaly )
+    anomaly.append( tm.anomaly )
+    anomalyProb.append( anomalyLikelihood )
 
     predictor.learn(count, tm.getActiveCells(), int(consumption / predictor_resolution))
 
@@ -192,7 +191,6 @@ def main(parameters=default_parameters, argv=None, verbose=True):
   # Calculate the predictive accuracy, Root-Mean-Squared
   accuracy         = {1: 0, 5: 0}
   accuracy_samples = {1: 0, 5: 0}
-  inputs = inputs[6:] #crop the first max prediction-steps inputs (as those don't have inferences)
 
   for idx, inp in enumerate(inputs):
     for n in predictions: # For each [N]umber of time steps ahead which was predicted.

src/htm/algorithms/SDRClassifier.cpp

@@ -47,8 +47,10 @@ void Classifier::initialize(const Real alpha)
 PDF Classifier::infer(const SDR & pattern) const {
   // Check input dimensions, or if this is the first time the Classifier is used and dimensions
   // are unset, return zeroes.
-  NTA_CHECK( dimensions_ != 0 )
-    << "Classifier: must call `learn` before `infer`.";
+  if( dimensions_ == 0 ) {
+    NTA_WARN << "Classifier: must call `learn` before `infer`.";
+    return PDF(numCategories_, std::nan("")); //empty array []
+  }
   NTA_ASSERT(pattern.size == dimensions_) << "Input SDR does not match previously seen size!";
 
   // Accumulate feed forward input.

src/htm/algorithms/SDRClassifier.hpp

@@ -128,6 +128,7 @@ class Classifier : public Serializable
    * @param pattern: The SDR containing the active input bits.
    * @returns: The Probablility Distribution Function (PDF) of the categories.
    *           This is indexed by the category label.
+   *           Or empty array ([]) if Classifier hasn't called learn() before. 
    */
   PDF infer(const SDR & pattern) const;