What is the correct method to construct a many-to-many LSTM model using Keras in Python?

Question

I am trying to make a 3 sequence many-to-many LSTM model, but I am confused about it's implementation in Keras. I searched on internet for examples of many-to-many models, but each website gives different method. That has confused me even more. What is the correct method of those? I want a model like this:

Some of the various methods I found were

Using encoder, decoder

from keras.layers import RepeatVector
from keras.layers import TimeDistributed

model = Sequential()

# encoder layer
model.add(LSTM(100, activation='relu', input_shape=(3, 1)))

# repeat vector
model.add(RepeatVector(3))

# decoder layer
model.add(LSTM(100, activation='relu', return_sequences=True))

model.add(TimeDistributed(Dense(1)))
model.compile(optimizer='adam', loss='mse')

Another with encoder, decoder

from keras.models import Model
from keras.layers import Input, LSTM, Dense

encoder_inputs = Input(shape=(None, 1))
encoder = LSTM(100, return_state=True)
encoder_outputs, state_h, state_c = encoder(encoder_inputs)
encoder_states = [state_h, state_c]


decoder_inputs = Input(shape=(None, 1))

decoder_lstm = LSTM(100, return_sequences=True, return_state=True)
decoder_outputs, _, _ = decoder_lstm(decoder_inputs,
                                     initial_state=encoder_states)
decoder_dense = Dense(num_decoder_tokens, activation='softmax')
decoder_outputs = decoder_dense(decoder_outputs)


model = Model([encoder_inputs, decoder_inputs], decoder_outputs)

model = Sequential()
model.add(LSTM(100,input_shape=(3,1),return_sequences=True))
model.add(TimeDistributed(Dense(2)))
model.compile(optimizer='adam', loss='mse')

model = Sequential()
model.add(LSTM(100,input_shape=(3,1),return_sequences=True))
model.compile(optimizer='adam', loss='mse')

Which one of these is the correct method? which one will give the model like the one I want?

score 0 · Answer 1 · answered Sep 01 '20 at 04:28

0

You have to mention your problem statement first.

1 and 2 are best for neural machine translation problems. While 2 is superior because it is considering return states in LSTM layer. 3 is also a good architecture where logic from input to output is simple. 4 is a very basic architecture becuase nth output in the output array has knowledge about [0 to n-1th input, not later ones] also no fully connected (Dense) layer so even moderate logic cannot be learned here.

answered Sep 01 '20 at 04:28

Sayan Dey

771
6
13

I posted the image of the model architecture. It is many to many with equal i/p o/p length – Shantanu Shinde Sep 01 '20 at 04:59
also, how do I apply the example in 2 for a sequence of 3 like I want? the one in 2 has sequence of 2 – Shantanu Shinde Sep 01 '20 at 05:01
Mention your use-case what do you want to do with your model, what's you dataset? – Sayan Dey Sep 01 '20 at 05:06
I want to get a certain vector o/p after a seq. of 3 embedded word vectors are given as i/p. Actually I think my model is wrong. there don't be o/p for 1st 2 lstms – Shantanu Shinde Sep 01 '20 at 05:18
So it's a language model? – Sayan Dey Sep 01 '20 at 05:21
are you sure you want 3 words as output only, or its variable – Sayan Dey Sep 01 '20 at 05:30
Go with 2 then, if that's a language model. – Sayan Dey Sep 01 '20 at 05:45
no, I just realized, I need a many to one model. any suggestions on that? – Shantanu Shinde Sep 01 '20 at 06:04
It doesn't work that way? You describe your dataset. You will be inputting words and outputting what? – Sayan Dey Sep 01 '20 at 06:08
I am inputting 3 words and outputting a single variable – Shantanu Shinde Sep 01 '20 at 07:31
What is that variable, is that a sentiment class, or some word? What is the actual problem statement? Model is selected on that basis only. – Sayan Dey Sep 01 '20 at 09:06
I am trying to predict the rgb value of color from it's name – Shantanu Shinde Sep 02 '20 at 04:09

What is the correct method to construct a many-to-many LSTM model using Keras in Python?

1 Answers1

Linked