improved edge correction for spike distance

Improvement following Eeros suggestions to use auxiliary spike at the edges
consistently with the corresponding corrected ISI intervals.

2 files changed, 46 insertions, 8 deletions

diff --git a/test/test_distance.py b/test/test_distance.py
index e45ac16..626b438 100644
--- a/test/test_distance.py
+++ b/test/test_distance.py
@@ -36,6 +36,7 @@ def test_isi():
     f = spk.isi_profile(t1, t2)
 
     # print("ISI: ", f.y)
+    print("ISI value:", expected_isi_val)
 
     assert_equal(f.x, expected_times)
     assert_array_almost_equal(f.y, expected_isi, decimal=15)
@@ -73,8 +74,19 @@ def test_spike():
 
     assert_equal(f.x, expected_times)
 
-    assert_almost_equal(f.avrg(), 1.6624149659863946e-01, decimal=15)
-    assert_almost_equal(f.y2[-1], 0.1394558, decimal=6)
+    # from SPIKY:
+    y_all = np.array([0.000000000000000000, 0.555555555555555580,
+                      0.222222222222222210, 0.305555555555555580,
+                      0.255102040816326536, 0.000000000000000000,
+                      0.000000000000000000, 0.255102040816326536,
+                      0.255102040816326536, 0.285714285714285698,
+                      0.285714285714285698, 0.285714285714285698])
+
+    #assert_array_almost_equal(f.y1, y_all[::2])
+    assert_array_almost_equal(f.y2, y_all[1::2])
+
+    assert_almost_equal(f.avrg(), 0.186309523809523814, decimal=15)
+    assert_equal(spk.spike_distance(t1, t2), f.avrg())
 
     t1 = SpikeTrain([0.2, 0.4, 0.6, 0.7], 1.0)
     t2 = SpikeTrain([0.3, 0.45, 0.8, 0.9, 0.95], 1.0)
@@ -99,6 +111,8 @@ def test_spike():
                              (expected_y1+expected_y2)/2)
     expected_spike_val /= (expected_times[-1]-expected_times[0])
 
+    print("SPIKE value:", expected_spike_val)
+
     f = spk.spike_profile(t1, t2)
 
     assert_equal(f.x, expected_times)
@@ -117,9 +131,14 @@ def test_spike():
     # for left and right values
     s1_r = np.array([0.1, (0.1*0.1+0.1*0.1)/0.2, 0.1, 0.0, 0.0, 0.0, 0.0])
     s1_l = np.array([0.1, (0.1*0.1+0.1*0.1)/0.2, 0.1, 0.0, 0.0, 0.0, 0.0])
-    s2_r = np.array([0.1*0.1/0.3, 0.1*0.3/0.3, 0.1*0.2/0.3,
+    # s2_r = np.array([0.1*0.1/0.3, 0.1*0.3/0.3, 0.1*0.2/0.3,
+    #                  0.0, 0.1, 0.0, 0.0])
+    # s2_l = np.array([0.1*0.1/0.3, 0.1*0.1/0.3, 0.1*0.2/0.3, 0.0,
+    #                  0.1, 0.0, 0.0])
+    # eero's edge correction:
+    s2_r = np.array([0.1, 0.1*0.3/0.3, 0.1*0.2/0.3,
                      0.0, 0.1, 0.0, 0.0])
-    s2_l = np.array([0.1*0.1/0.3, 0.1*0.1/0.3, 0.1*0.2/0.3, 0.0,
+    s2_l = np.array([0.1, 0.1*0.3/0.3, 0.1*0.2/0.3, 0.0,
                      0.1, 0.0, 0.0])
     isi1 = np.array([0.2, 0.2, 0.2, 0.2, 0.2, 0.4])
     isi2 = np.array([0.3, 0.3, 0.3, 0.1, 0.1, 0.4])
@@ -345,7 +364,7 @@ def test_regression_spiky():
     assert_equal(isi_profile.y, 0.1/1.1 * np.ones_like(isi_profile.y))
 
     spike_dist = spk.spike_distance(st1, st2)
-    assert_equal(spike_dist, 2.1105878248735391e-01)
+    assert_equal(spike_dist, 0.211058782487353908)
 
     spike_sync = spk.spike_sync(st1, st2)
     assert_equal(spike_sync, 8.6956521739130432e-01)
@@ -363,12 +382,31 @@ def test_regression_spiky():
 
     spike_dist = spk.spike_distance_multi(spike_trains)
     # get the full precision from SPIKY
-    assert_almost_equal(spike_dist, 2.4432433330596512e-01, decimal=15)
+    assert_almost_equal(spike_dist, 0.25188056475463755, decimal=15)
 
     spike_sync = spk.spike_sync_multi(spike_trains)
     # get the full precision from SPIKY
     assert_equal(spike_sync, 0.7183531505298066)
 
+    # Eero's edge correction example
+    st1 = SpikeTrain([0.5, 1.5, 2.5], 6.0)
+    st2 = SpikeTrain([3.5, 4.5, 5.5], 6.0)
+
+    f = spk.spike_profile(st1, st2)
+
+    expected_times = np.array([0.0, 0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 6.0])
+    y_all = np.array([0.271604938271605, 0.271604938271605, 0.271604938271605,
+                      0.617283950617284, 0.617283950617284, 0.444444444444444,
+                      0.285714285714286, 0.285714285714286, 0.444444444444444,
+                      0.617283950617284, 0.617283950617284, 0.271604938271605,
+                      0.271604938271605, 0.271604938271605])
+    expected_y1 = y_all[::2]
+    expected_y2 = y_all[1::2]
+
+    assert_equal(f.x, expected_times)
+    assert_array_almost_equal(f.y1, expected_y1, decimal=14)
+    assert_array_almost_equal(f.y2, expected_y2, decimal=14)
+
 
 def test_multi_variate_subsets():
     spike_trains = spk.load_spike_trains_from_txt("test/PySpike_testdata.txt",
diff --git a/test/test_empty.py b/test/test_empty.py
index af7fb36..5a0042f 100644
--- a/test/test_empty.py
+++ b/test/test_empty.py
@@ -70,8 +70,8 @@ def test_spike_empty():
     st1 = SpikeTrain([], edges=(0.0, 1.0))
     st2 = SpikeTrain([0.4, ], edges=(0.0, 1.0))
     d = spk.spike_distance(st1, st2)
-    assert_almost_equal(d, 0.4*0.4*1.0/(0.4+1.0)**2 + 0.6*0.4*1.0/(0.6+1.0)**2,
-                        decimal=15)
+    d_expect = 0.4*0.4*1.0/(0.4+1.0)**2 + 0.6*0.4*1.0/(0.6+1.0)**2
+    assert_almost_equal(d, d_expect, decimal=15)
     prof = spk.spike_profile(st1, st2)
     assert_equal(d, prof.avrg())
     assert_array_equal(prof.x, [0.0, 0.4, 1.0])