-Add plotting for the first two eigenvectors of

STC matrix.
-STC does not work properly, test it further
and increase efficiency.

LNP_model

@@ -15,15 +15,15 @@ t = np.arange(0, total_frames*dt, dt)  # Time vector
 filter_time = .6  # The longest feature RGCs respond to is ~600ms
 filter_length = int(filter_time/dt)  # Filter is filter_length frames long
 
-cweight = 1  # The weight of combination for the two filters
+cweight = .5  # The weight of combination for the two filters
 
 
 def make_noise():  # Generate gaussian noise for stimulus
     return np.random.normal(0, 9, total_frames)
-# stimulus=make_noise()
+stimulus = make_noise()
 
 filter_index1 = 2   # Change filter type here
-filter_index2 = 1
+filter_index2 = 3
 
 
 def linear_filter(t, filter_index):    # Define filter according to choice

stc.py

@@ -22,9 +22,17 @@ def stc(spikes, stimulus, filter_length, sta_temp):
             # Snippets are inverted before being added
             snpta = np.matrix(snippet-sta_temp)
             covariance = covariance+(snpta.T*snpta)*spikes[i]
-    return covariance/(sum(spikes)+filter_length-1)
+    return covariance/(sum(spikes)*filter_length-1)
 
 recovered_stc = stc(spikes, stimulus, filter_length,
                     sta(spikes, stimulus, filter_length))
 runtime = str(datetime.now()-execution_timer).split('.')[0]
 print('Duration: {}'.format(runtime))
+
+
+w,v = np.linalg.eig(recovered_stc)
+# column v[:,i] is the eigenvector corresponding to the eigenvalue w[i]
+plt.plot(w, 'o', markersize=2)
+plt.show()
+plt.plot(v[:,1],'b')
+plt.plot(v[:,2],'g')