LSTM followed by the middle pool

Question

I am using Keras 1.0. My problem is identical to this ( How to implement a middle layer pool in Keras ), but the answer there seems to be not enough for me.

I want to implement this network:

The following code does not work:

sequence = Input(shape=(max_sent_len,), dtype='int32') embedded = Embedding(vocab_size, word_embedding_size)(sequence) lstm = LSTM(hidden_state_size, activation='sigmoid', inner_activation='hard_sigmoid', return_sequences=True)(embedded) pool = AveragePooling1D()(lstm) output = Dense(1, activation='sigmoid')(pool) 

If I do not set return_sequences=True , I get this error when I call AveragePooling1D() :

 Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/PATH/keras/engine/topology.py", line 462, in __call__ self.assert_input_compatibility(x) File "/PATH/keras/engine/topology.py", line 382, in assert_input_compatibility str(K.ndim(x))) Exception: ('Input 0 is incompatible with layer averagepooling1d_6: expected ndim=3', ' found ndim=2') 

Otherwise, I get this error when I call Dense() :

 Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/PATH/keras/engine/topology.py", line 456, in __call__ self.build(input_shapes[0]) File "/fs/clip-arqat/mossaab/trec/liveqa/cmu/venv/lib/python2.7/site-packages/keras/layers/core.py", line 512, in build assert len(input_shape) == 2 AssertionError 

Answer 1

Adding TimeDistributed(Dense(1)) helped:

 sequence = Input(shape=(max_sent_len,), dtype='int32') embedded = Embedding(vocab_size, word_embedding_size)(sequence) lstm = LSTM(hidden_state_size, activation='sigmoid', inner_activation='hard_sigmoid', return_sequences=True)(embedded) distributed = TimeDistributed(Dense(1))(lstm) pool = AveragePooling1D()(distributed) output = Dense(1, activation='sigmoid')(pool) 

Answer 2

Please try this (I hope this solves your problem :)):

http://keras.io/layers/core/#timedistributedmerge

Answer 3

I think the accepted answer is basically wrong. The solution was discovered at: https://github.com/fchollet/keras/issues/2151 However, it works only for the backend. I modified the code to support both anano and tensor flow.

 from keras.engine.topology import Layer, InputSpec from keras import backend as T class TemporalMeanPooling(Layer): """ This is a custom Keras layer. This pooling layer accepts the temporal sequence output by a recurrent layer and performs temporal pooling, looking at only the non-masked portion of the sequence. The pooling layer converts the entire variable-length hidden vector sequence into a single hidden vector, and then feeds its output to the Dense layer. input shape: (nb_samples, nb_timesteps, nb_features) output shape: (nb_samples, nb_features) """ def __init__(self, **kwargs): super(TemporalMeanPooling, self).__init__(**kwargs) self.supports_masking = True self.input_spec = [InputSpec(ndim=3)] def get_output_shape_for(self, input_shape): return (input_shape[0], input_shape[2]) def call(self, x, mask=None): #mask: (nb_samples, nb_timesteps) if mask is None: mask = T.mean(T.ones_like(x), axis=-1) ssum = T.sum(x,axis=-2) #(nb_samples, np_features) mask = T.cast(mask,T.floatx()) rcnt = T.sum(mask,axis=-1,keepdims=True) #(nb_samples) return ssum/rcnt #return rcnt def compute_mask(self, input, mask): return None 

Answer 4

I just tried to implement the same model as the original poster, and I am using Keras 2.0.3 . The middle pool after LSTM worked when I used GlobalAveragePooling1D , just make sure return_sequences=True in the LSTM layer. Give it a try!

Answer 5

Thank you, I also answer the question, but I think that the TimeDistributed layer does not work the way you want, you can try the Luke Guye TemporalMeanPooling layer, it works for me. Here is an example:

 sequence = Input(shape=(max_sent_len,), dtype='int32') embedded = Embedding(vocab_size, word_embedding_size)(sequence) lstm = LSTM(hidden_state_size, return_sequences=True)(embedded) pool = TemporalMeanPooling()(lstm) output = Dense(1, activation='sigmoid')(pool)