feat(sttc): more efficient calculation

Calculating the spike time tiling matrix using `spike_time_tilings()`
used to duplicate the work of calculating the constants TA and TB for
each of the units. It's not as useful as I expected, as it only seems
to be about 10% faster, but I've already done and tested it.

braingeneers/analysis/analysis.py

@@ -833,15 +833,21 @@ def concatenate_spike_data(self, sd):
 
 
     def spike_time_tilings(self, delt=20):
-        '''
+        """
         Compute the full spike time tiling coefficient matrix.
-        '''
+        """
+        T = self.length
+        ts = [_sttc_ta(ts, delt, T) / T for ts in self.train]
+
         ret = np.diag(np.ones(self.N))
         for i in range(self.N):
             for j in range(i + 1, self.N):
-                ret[i, j] = ret[j, i] = self.spike_time_tiling(i, j, delt)
+                ret[i, j] = ret[j, i] = _spike_time_tiling(
+                    self.train[i], self.train[j], ts[i], ts[j], delt
+                )
         return ret
 
+
     def spike_time_tiling(self, i, j, delt=20):
         '''
         Calculate the spike time tiling coefficient between two units within
@@ -1010,7 +1016,11 @@ def spike_time_tiling(tA, tB, delt=20, length=None):
 
     TA = _sttc_ta(tA, delt, length) / length
     TB = _sttc_ta(tB, delt, length) / length
+    return _spike_time_tiling(tA, tB, TA, TB, delt)
+
 
+def _spike_time_tiling(tA, tB, TA, TB, delt):
+    "Internal helper method for the second half of STTC calculation."
     PA = _sttc_na(tA, tB, delt) / len(tA)
     PB = _sttc_na(tB, tA, delt) / len(tB)
 

braingeneers/analysis/analysis_test.py

@@ -359,6 +359,7 @@ def test_spike_time_tiling_coefficient(self):
         self.assertEqual(sttc[0, 1], sttc[1, 0])
         self.assertEqual(sttc[0, 0], 1.0)
         self.assertEqual(sttc[1, 1], 1.0)
+        self.assertEqual(sttc[0, 1], foo.spike_time_tiling(0, 1, 1))
 
         # Default arguments, inferred value of tmax.
         tmax = max(foo.train[0].ptp(), foo.train[1].ptp())