Update model

channelCS · May 25, 2018 · 88eface · 88eface
1 parent 98a70a4
commit 88eface
Show file tree

Hide file tree

Showing 3 changed files with 26 additions and 63 deletions.
diff --git a/aud_model.py b/aud_model.py
@@ -107,28 +107,6 @@ def prepare_model(self):
             raise ValueError("Could not find model {}".format(self.model))
         return lrmodel
 
-class Static_Model:
-    def __init__(self,input_neurons,cross_validation,
-        dimx,dimy,num_classes,
-        nb_filter,filter_length,
-        model):
-        if model is None:
-            raise ValueError("No model passed")
-        self.cross_validation=cross_validation
-        self.input_neurons=input_neurons
-        self.model=model
-        self.nb_filter = nb_filter
-        self.filter_length =filter_length
-        self.dimx = dimx
-        self.dimy = dimy
-        self.num_classes=num_classes
-
-    def prepare_model(self):
-        if self.model=='CHOU':
-            lrmodel=M.conv_deconv_chou(dimx=self.dimx,dimy=self.dimy,nb_filter=self.nb_filter,num_classes=self.num_classes)
-        else:
-            raise ValueError("Could not find model {}".format(self.model))
-        return lrmodel
 
 class Dynamic_Model:
     def __init__(self,model,num_classes,dimx,dimy,acts,**kwargs):

diff --git a/aud_utils.py b/aud_utils.py
@@ -204,7 +204,7 @@ def mat_3d_to_nd(model, X):
     [batch_num, dimx, dimy]= X.shape 
     two_d   = ['DNN']
     three_d = ['RNN']
-    four_d  = ['CNN', 'CHOU', 'CRNN', 'FCRNN', 'CBRNN', 'MultiCNN', 'TCNN','ACRNN', 'MultiACRNN']
+    four_d  = ['CNN', 'CRNN', 'FCRNN', 'CBRNN', 'MultiCNN', 'TCNN','ACRNN']
     if model in two_d:
         X = X.reshape(batch_num, dimx*dimy)    
     elif model in three_d:

diff --git a/model_description.py b/model_description.py
@@ -7,7 +7,7 @@
 from keras.models import Model
 from keras.layers import Dense, Dropout, Flatten, Input
 from keras.layers import Conv2D, Conv2DTranspose, merge, Merge
-from keras.layers import BatchNormalization, Lambda,Activation,concatenate,RepeatVector,dot
+from keras.layers import BatchNormalization, Lambda,Activation,Concatenate,RepeatVector,dot
 from keras.layers import LSTM, GRU, Reshape, Bidirectional, Permute,TimeDistributed
 from keras.layers import MaxPooling2D, AveragePooling2D, GlobalMaxPooling1D, GlobalAveragePooling2D
 from keras.layers.merge import Multiply
@@ -152,15 +152,18 @@ def cnn(dimx,dimy,num_classes,**kwargs):
     act2           = kwargs['kwargs'].get('act2','relu')
     act3           = kwargs['kwargs'].get('act3','softmax')
     dropout        = kwargs['kwargs'].get('dropout',0.1)
-    nb_filter      = kwargs['kwargs'].get('nb_filter',[])
+    nb_filter      = kwargs['kwargs'].get('nb_filter',128)
     filter_length  = kwargs['kwargs'].get('filter_length',3)
-    pool_size      = kwargs['kwargs'].get('pool_size',[])
+    pool_size      = kwargs['kwargs'].get('pool_size',2)
     print_sum      = kwargs['kwargs'].get('print_sum',False)
 
     loss          = kwargs['kwargs'].get('loss','categorical_crossentropy')
     optimizer     = kwargs['kwargs'].get('optimizer','adam')
     metrics       = kwargs['kwargs'].get('metrics','accuracy')
-
+    if type(nb_filter) is int:
+        nb_filter = [nb_filter] * 2
+    if type(pool_size) is int:
+        pool_size = [pool_size] * 2
     print "Model CNN"
     print "Activation 1 {} 2 {} 3 {}".format(act1,act2,act3)
     print "Neurons {} Dropout {}".format(input_neurons,dropout)
@@ -550,14 +553,16 @@ def transpose_cnn(dimx,dimy,num_classes,**kwargs):
     act1          = kwargs['kwargs'].get('act1','tanh')
     act2          = kwargs['kwargs'].get('act2','tanh')
     act3          = kwargs['kwargs'].get('act3','sigmoid')
-    nb_filter      = kwargs['kwargs'].get('nb_filter',[])
+    nb_filter      = kwargs['kwargs'].get('nb_filter',128)
     pool_size      = kwargs['kwargs'].get('pool_size',(1,2))
     dropout        = kwargs['kwargs'].get('dropout',0.1)
     print_sum      = kwargs['kwargs'].get('print_sum',False)
 
     loss          = kwargs['kwargs'].get('loss','binary_crossentropy')
     optimizer     = kwargs['kwargs'].get('optimizer','adam')
     metrics       = kwargs['kwargs'].get('metrics','mse')
+    if type(nb_filter) is int:
+        nb_filter = [nb_filter] * 2
     inpx = Input(shape=(1,dimx,dimy),name='inpx')
     x = Conv2D(filters=nb_filter[0],
                kernel_size=5,
@@ -603,47 +608,27 @@ def seq2seq(dimx,dimy,num_classes,**kwargs):
     print "seq2seq_lstm"
 
     ## encoder
-    encoder_input = Input(shape=(dimx,dimy))
-
-    encoder=Bidirectional(LSTM(32,return_state=True))# Returns list of nos. of output states
-    encoder_outputs, forward_h, forward_c, backward_h, backward_c = encoder(encoder_input)
+    x = Input(shape=(dimx,dimy))
+    # Encoder
+    encoder=Bidirectional(LSTM(32,return_state=True))
+    encoder_outputs, forward_h, forward_c, backward_h, backward_c = encoder(x)
     state_h = Concatenate(axis=1)([forward_h, backward_h])
     state_c = Concatenate(axis=1)([forward_c, backward_c])
     encoder_states = [state_h, state_c]
+    #hidden_1 = Dense(128, activation='relu')(encoder_outputs)
+    #h = Dense(64, activation='relu')(hidden_1)
 
-#    a,b = kr(encoder_outputs)
-#    x = Reshape((b,1))(encoder_outputs)
-
-
-    ## decoder
-    decoder_input = Input(shape=(dimx,dimy), name='main_input')
-
+    # Decoder
+    y = Input(shape=(dimx,dimy))
     decoder_lstm = LSTM(64, return_sequences=True, return_state=True)
-    decoder_outputs, _, _ = decoder_lstm(decoder_input,
+    decoder_outputs, _, _ = decoder_lstm(y,
                                          initial_state=encoder_states)
-    #h=Flatten()(decoder_outputs)
-    decoder_dense = Dense(40, activation='softmax')
-    decoder_outputs=decoder_dense(decoder_outputs)
-    model = Model([encoder_input, decoder_input], decoder_outputs)
-    model.summary()
-    model.compile(loss='categorical_crossentropy',
-			  optimizer='adam',
-			  metrics=['accuracy'])
-#
-#    ## encoder model 
-#    encoder_model = Model(encoder_input, encoder_states)
-#    
-#    ##decoder model
-#    
-#    decoder_state_input_h = Input(shape=(64,))
-#    decoder_state_input_c = Input(shape=(64,))
-#    decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]
-#    decoder_outputs, state_h, state_c = decoder_lstm(
-#    decoder_input, initial_state=decoder_states_inputs)
-#    decoder_states = [state_h, state_c]
-#    decoder_outputs = decoder_dense(decoder_outputs)
-#    decoder_model = Model( [decoder_input] + decoder_states_inputs,[decoder_outputs] + decoder_states)
-#  
+    #hidden_2 = Dense(128, activation='relu')(decoder_outputs)
+    r = Dense(40, activation='sigmoid')(decoder_outputs)
+
+    model = Model([x,y], r)
+    model.compile(optimizer='adam', loss='categorical_crossentropy',metrics=['accuracy'])
+
     return model