comparison cnmf.py @ 11:915c849b17ea

added arg in cnmf script to choose to run standard nmf

     nmf_mdl.factorize(niter=niter)
     F = np.asarray(nmf_mdl.W)
     G = np.asarray(nmf_mdl.H)
     return F, G
 
-
+def nmf(S, rank, nither=500):
+	nmf_mdl = pymf.NMF(S, num_bases=rank, niter=nither)
+	nmf_mdl.factorize()
+    F = np.asarray(nmf_mdl.W)
+    G = np.asarray(nmf_mdl.H)
+    return F, G
+	
+	
 def most_frequent(x):
     """Returns the most frequent value in x."""
     return np.argmax(np.bincount(x))
 
 

     G = np.sum(G, axis=1)
     G = SegUtil.median_filter(G[:, np.newaxis], R)
     return G.flatten()
 
 
-def segmentation(X, rank=4, R=15, h=8, niter=300):
+def segmentation(X, rank=4, R=15, h=8, niter=300, CNMF=True):
     """
     Gets the segmentation (boundaries and labels) from the factorization
     matrices.
 
     Parameters

         Size of the median filter for activation matrix
     niter: int
         Number of iterations for k-means
     bound_idxs : list
         Use previously found boundaries (None to detect them)
-
+	CNMF : bool
+		If True, use CNMF; otherwise use NMF
+		
     Returns
     -------
     bounds_idx: np.array
         Bound indeces found
     labels: np.array

     # Find non filtered boundaries
     bound_idxs = None
     while True:
         if bound_idxs is None:
             try:
-                F, G = cnmf(X, rank, niter=niter)
+                if CNMF: F, G = cnmf(X, rank, niter=niter)
+				else: F, G = nmf(X, rank, niter=niter)
             except:
                 return np.empty(0), [1]
 
             # Filter G
             G = filter_activation_matrix(G.T, R)