--- a/pylearn/sandbox/scan_inputs_groups.py	Thu May 14 20:00:39 2009 -0400
+++ b/pylearn/sandbox/scan_inputs_groups.py	Thu May 14 20:01:04 2009 -0400
@@ -105,10 +105,10 @@
 			raise NotImplementedError('index superior to weight list length',idx_list)
 		for i in range(len(args[1])):
 			if (args[1][i].shape)[0] != batchsize:
-				raise NotImplementedError('different batchsize in the inputs list',args[1][i])
+				raise NotImplementedError('different batchsize in the inputs list',args[1][i].shape)
 		for i in range(len(args)-2):
 			if (args[2+i].shape)[1] != n_hid:
-				raise NotImplementedError('different length of hidden in the weights list',args[2+i])
+				raise NotImplementedError('different length of hidden in the weights list',args[2+i].shape)
 		
 		for i in range(len(idx_list)):
 			if idx_list[i]>0:
@@ -168,10 +168,10 @@
 			raise NotImplementedError('index superior to weight list length',idx_list)
 		for i in range(len(args[1])):
 			if (args[1][i].shape)[0] != batchsize:
-				raise NotImplementedError('different batchsize in the inputs list',args[1][i])
+				raise NotImplementedError('different batchsize in the inputs list',args[1][i].shape)
 		for i in range(len(args)-3):
 			if (args[2+i].shape)[1] != n_hid:
-				raise NotImplementedError('different length of hidden in the weights list',args[2+i])
+				raise NotImplementedError('different length of hidden in the weights list',args[2+i].shape)
 		
 		zcalc = [False for i in range(len(args)-3)]
 		
@@ -234,7 +234,7 @@
 			raise NotImplementedError('size of index different of inputs list size',idx_list)
 		for i in range(len(args)-3):
 			if (args[3+i].shape)[0] != n_hid:
-				raise NotImplementedError('different length of hidden in the weights list',args[3+i])
+				raise NotImplementedError('different length of hidden in the weights list',args[3+i].shape)
 		
 		zcalc = [False for i in idx_list]
 		z[0] = [None for i in idx_list]
@@ -304,7 +304,7 @@
 			raise NotImplementedError('size of index different of inputs list size',idx_list)
 		for i in range(len(args)-4):
 			if (args[3+i].shape)[0] != n_hid:
-				raise NotImplementedError('different length of hidden in the weights list',args[3+i])
+				raise NotImplementedError('different length of hidden in the weights list',args[3+i].shape)
 		
 		zidx=numpy.zeros((len(idx_list)+1))
 		
@@ -377,7 +377,7 @@
 		if len(idx_list) != len(inputs_list) :
 			raise NotImplementedError('size of index different of inputs list size',idx_list)
 		
-		y[0] = [-i if (i>0 and not(self.R.noisify_group_bool(noise_level_group))) else i for i in idx_list]
+		y[0] = numpy.asarray([-i if (i>0 and not(self.R.noisify_group_bool(noise_level_group))) else i for i in idx_list])
 		z[0] = [(self.R.noisify_bit(inputs_list[i],noise_level_bit) if y[0][i]>0 else numpy.zeros((inputs_list[i].shape)))\
 				for i in range(len(inputs_list))]
 	
@@ -386,7 +386,7 @@
 	
 	
 	def __hash__(self):
-		return hash(ScanNoise)^hash(self.seed)^hash(self.R)^12254
+		return hash(ScanNoise)^hash(self.seed)^hash(self.R.rand)^12254
 	
 	def __str__(self):
 		return "ScanNoise"
@@ -500,3 +500,50 @@
 	
 	def __str__(self):
 		return "ScanBiasDecGrad"
+
+# Mask construction------------------------------------
+class ScanMask(Op):
+	"""This Op takes an index list (as tensor.ivector) and a list of weigths.
+	It will construct a list of T.iscalar representing the Mask
+	to do the correct regularisation on the weigths"""
+	def __init__(self,encbool=True):
+		self.encbool = encbool
+	
+	def make_node(self, idx_list, weights_list):
+		idx_list = Checkidx_list(idx_list)
+		weights_list = Checkweights_list(weights_list)
+		return Apply(self, [idx_list] + weights_list, [T.iscalar() for i in range(len(weights_list))])
+	
+	def perform(self, node, args, z):
+		if self.encbool:
+			idx_list = args[0][args[0]>0]
+			dim = 1
+		else:
+			idx_list = abs(args[0][args[0] != 0])
+			dim = 0
+		n_hid = args[1].shape[dim]
+		
+		if max(idx_list) >= (len(args)-1)+1 :
+			raise NotImplementedError('index superior to weights list length',idx_listdec)
+		for i in range(len(args)-1):
+			if args[1+i].shape[dim] != n_hid:
+				raise NotImplementedError('different length of hidden in the encoding weights list',args[1+i].shape)
+		
+		for i in range(len(args[1:])):
+			z[i][0] = numpy.asarray((idx_list == i+1).sum(),dtype='int32')
+	
+	def __hash__(self):
+		return hash(ScanMask)^hash(self.encbool)^11447
+	
+	def grad(self,args,gz):
+		return [None] * len(args)
+	
+	def __str__(self):
+		if self.encbool:
+			string = "Enc"
+		else:
+			string = "Dec"
+		return "ScanMask" + string
+
+scanmaskenc=ScanMask(True)
+scanmaskdec=ScanMask(False)
\ No newline at end of file