Revert "Connections:computeActivity returns connected counts per segment"

This reverts commit 8d38633.

Author: breznak

bindings/py/cpp_src/bindings/algorithms/py_Connections.cpp

@@ -101,7 +101,7 @@ R"(Compatibility Warning: This classes API is unstable and may change without wa
                 [](void *dataPtr) {
                     delete reinterpret_cast<std::vector<SynapseIdx>*>(dataPtr); });
             // Call the C++ method.
-            *activeConnectedSynapses = self.computeActivity(activePresynapticCells.getSparse());
+            self.computeActivity(*activeConnectedSynapses, activePresynapticCells.getSparse());
             // Wrap vector in numpy array.
             return py::array(activeConnectedSynapses->size(),
                              activeConnectedSynapses->data(),
@@ -124,7 +124,7 @@ R"(Returns numActiveConnectedSynapsesForSegment)");
                 [](void *dataPtr) {
                     delete reinterpret_cast<std::vector<SynapseIdx>*>(dataPtr); });
             // Call the C++ method.
-            *activeConnectedSynapses = self.computeActivity(*activePotentialSynapses,
+            self.computeActivity(*activeConnectedSynapses, *activePotentialSynapses,
                                             activePresynapticCells.getSparse());
             // Wrap vector in numpy array.
             return py::make_tuple(

src/htm/algorithms/Connections.cpp

@@ -391,10 +391,11 @@ void Connections::reset()
   currentUpdates_.clear();
 }
 
-vector<SynapseIdx> Connections::computeActivity(
+void Connections::computeActivity(
+    vector<SynapseIdx> &numActiveConnectedSynapsesForSegment,
     const vector<CellIdx> &activePresynapticCells)
 {
-  vector<SynapseIdx> numActiveConnectedSynapsesForSegment(segments_.size(), 0);
+  NTA_ASSERT(numActiveConnectedSynapsesForSegment.size() == segments_.size());
 
   if( timeseries_ ) {
     // Before each cycle of computation move the currentUpdates to the previous
@@ -411,30 +412,31 @@ vector<SynapseIdx> Connections::computeActivity(
       }
     }
   }
-  return numActiveConnectedSynapsesForSegment;
 }
 
-vector<SynapseIdx> Connections::computeActivity(
+void Connections::computeActivity(
+    vector<SynapseIdx> &numActiveConnectedSynapsesForSegment,
     vector<SynapseIdx> &numActivePotentialSynapsesForSegment,
     const vector<CellIdx> &activePresynapticCells) {
+  NTA_ASSERT(numActiveConnectedSynapsesForSegment.size() == segments_.size());
   NTA_ASSERT(numActivePotentialSynapsesForSegment.size() == segments_.size());
 
   // Iterate through all connected synapses.
-  const vector<SynapseIdx> numActiveConnectedSynapsesForSegment = computeActivity( activePresynapticCells );
+  computeActivity(
+      numActiveConnectedSynapsesForSegment,
+      activePresynapticCells );
 
   // Iterate through all potential synapses.
   std::copy( numActiveConnectedSynapsesForSegment.begin(),
              numActiveConnectedSynapsesForSegment.end(),
              numActivePotentialSynapsesForSegment.begin());
-
   for (const auto& cell : activePresynapticCells) {
     if (potentialSegmentsForPresynapticCell_.count(cell)) {
       for(const auto& segment : potentialSegmentsForPresynapticCell_.at(cell)) {
         ++numActivePotentialSynapsesForSegment[segment];
       }
     }
   }
-  return numActiveConnectedSynapsesForSegment;
 }
 
 

src/htm/algorithms/Connections.hpp

@@ -389,20 +389,21 @@ class Connections : public Serializable
    * preinitialized with the length returned by
    * getSegmentFlatVectorLength().
    *
-   * @param (optional) numActivePotentialSynapsesForSegment
+   * @param numActiveConnectedSynapsesForSegment
+   * An output vector for active connected synapse counts per segment.
+   *
+   * @param numActivePotentialSynapsesForSegment
    * An output vector for active potential synapse counts per segment.
    *
    * @param activePresynapticCells
    * Active cells in the input.
-   *
-   * @return numActiveConnectedSynapsesForSegment
-   * An output vector for active connected synapse counts per segment.
    */
-  std::vector<SynapseIdx> computeActivity(
+  void computeActivity(std::vector<SynapseIdx> &numActiveConnectedSynapsesForSegment,
                        std::vector<SynapseIdx> &numActivePotentialSynapsesForSegment,
                        const std::vector<CellIdx> &activePresynapticCells);
 
-  std::vector<SynapseIdx> computeActivity(const std::vector<CellIdx> &activePresynapticCells);
+  void computeActivity(std::vector<SynapseIdx> &numActiveConnectedSynapsesForSegment,
+                       const std::vector<CellIdx> &activePresynapticCells);
 
   /**
    * The primary method in charge of learning.   Adapts the permanence values of

src/htm/algorithms/SpatialPooler.cpp

@@ -814,7 +814,8 @@ void SpatialPooler::updateBookeepingVars_(bool learn) {
 
 void SpatialPooler::calculateOverlap_(const SDR &input,
                                       vector<SynapseIdx> &overlaps) {
-  overlaps = connections_.computeActivity(input.getSparse());
+  overlaps.assign( numColumns_, 0 );
+  connections_.computeActivity(overlaps, input.getSparse());
 }
 
 

src/htm/algorithms/SpatialPooler.hpp

@@ -886,7 +886,9 @@ class SpatialPooler : public Serializable
 
      The overlap of a column is the number of synapses for that column
      that are connected (permanence value is greater than
-     '_synPermConnected') to input bits which are turned on.
+     '_synPermConnected') to input bits which are turned on. The
+     implementation takes advantage of the SparseBinaryMatrix class to
+     perform this calculation efficiently.
 
      @param inputVector
      a int array of 0's and 1's that comprises the input to the spatial

src/htm/algorithms/TemporalMemory.cpp

@@ -512,8 +512,9 @@ void TemporalMemory::activateDendrites(const bool learn,
 
   const size_t length = connections.segmentFlatListLength();
 
+  numActiveConnectedSynapsesForSegment_.assign(length, 0);
   numActivePotentialSynapsesForSegment_.assign(length, 0);
-  numActiveConnectedSynapsesForSegment_ = connections.computeActivity(
+  connections.computeActivity(numActiveConnectedSynapsesForSegment_,
                               numActivePotentialSynapsesForSegment_,
                               activeCells_);
 

src/test/unit/algorithms/ConnectionsTest.cpp

@@ -71,9 +71,12 @@ void setupSampleConnections(Connections &connections) {
 void computeSampleActivity(Connections &connections) {
   vector<UInt32> input = {50, 52, 53, 80, 81, 82, 150, 151};
 
+  vector<SynapseIdx> numActiveConnectedSynapsesForSegment(
+      connections.segmentFlatListLength(), 0);
   vector<SynapseIdx> numActivePotentialSynapsesForSegment(
       connections.segmentFlatListLength(), 0);
-  vector<SynapseIdx> numActiveConnectedSynapsesForSegment = connections.computeActivity(numActivePotentialSynapsesForSegment, input);
+  connections.computeActivity(numActiveConnectedSynapsesForSegment,
+                              numActivePotentialSynapsesForSegment, input);
 }
 
 /**
@@ -150,9 +153,11 @@ TEST(ConnectionsTest, testDestroySegment) {
   ASSERT_EQ(3ul, connections.numSegments());
   ASSERT_EQ(0ul, connections.numSynapses());
 
+  vector<SynapseIdx> numActiveConnectedSynapsesForSegment(
+      connections.segmentFlatListLength(), 0);
   vector<SynapseIdx> numActivePotentialSynapsesForSegment(
       connections.segmentFlatListLength(), 0);
-  vector<SynapseIdx> numActiveConnectedSynapsesForSegment = connections.computeActivity(
+  connections.computeActivity(numActiveConnectedSynapsesForSegment,
                               numActivePotentialSynapsesForSegment,
                               {80, 81, 82});
 
@@ -179,9 +184,11 @@ TEST(ConnectionsTest, testDestroySynapse) {
   ASSERT_EQ(2ul, connections.numSynapses());
   ASSERT_EQ(2ul, connections.synapsesForSegment(segment).size());
 
+  vector<SynapseIdx> numActiveConnectedSynapsesForSegment(
+      connections.segmentFlatListLength(), 0);
   vector<SynapseIdx> numActivePotentialSynapsesForSegment(
       connections.segmentFlatListLength(), 0);
-  vector<SynapseIdx> numActiveConnectedSynapsesForSegment = connections.computeActivity(
+  connections.computeActivity(numActiveConnectedSynapsesForSegment,
                               numActivePotentialSynapsesForSegment,
                               {80, 81, 82});
 
@@ -335,9 +342,11 @@ TEST(ConnectionsTest, testComputeActivity) {
 
   vector<UInt32> input = {50, 52, 53, 80, 81, 82, 150, 151};
 
+  vector<SynapseIdx> numActiveConnectedSynapsesForSegment(
+      connections.segmentFlatListLength(), 0);
   vector<SynapseIdx> numActivePotentialSynapsesForSegment(
       connections.segmentFlatListLength(), 0);
-  vector<SynapseIdx> numActiveConnectedSynapsesForSegment = connections.computeActivity(
+  connections.computeActivity(numActiveConnectedSynapsesForSegment,
                               numActivePotentialSynapsesForSegment, input);
 
   ASSERT_EQ(1ul, numActiveConnectedSynapsesForSegment[segment1_1]);
@@ -705,7 +714,7 @@ TEST(ConnectionsTest, testTimeseries) {
   presyn.randomize( 0.5f );
   vector<SynapseIdx> output( 1u );
   for( int i = 0; i < 10; i++ ) {
-    output = C.computeActivity( presyn.getSparse() );
+    C.computeActivity( output, presyn.getSparse() );
     C.adaptSegment( seg, presyn, 0.1f, 0.1f );
   }
   // Check that the synapse permanences did not saturate.
@@ -718,7 +727,7 @@ TEST(ConnectionsTest, testTimeseries) {
   // effectively turns off the timeseries parameter.
   for( int i = 0; i < 10; i++ ) {
     C.reset();
-    output = C.computeActivity( presyn.getSparse() );
+    C.computeActivity( output, presyn.getSparse() );
     C.adaptSegment( seg, presyn, 0.1f, 0.1f );
   }
   // Check that the synapse permanences staturated.  This is the failure