CODEWITHSUNDEEP
lstmrecurrent-neural-network
cadyT
cadyT

Reputation: 77

Keras implementation for RNN

I have coded LSTM as below. Now I would like to compare the performance of the RNN and LSTM. Actually, I know LSTM is a type of RNN. But how can I take the results from RNN on Keras? I could not find a proper RNN code example on Keras.

model = Sequential()
model.add(LSTM(15, input_shape=(max_fixation_length, feature_size,), return_sequences=True))
model.add(Flatten())
model.add(Dense(1, activation='sigmoid'))
# Compile model
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])enter code here

Also, I have read this link Keras simple RNN implementation . However, I could not run it. Because Keras has given this error "undefined name 'SimpleRNN'"

Upvotes: 0

Views: 592

Answers (2)

EverythingNeedToBeKnown
EverythingNeedToBeKnown

Reputation: 115

Simply first you need to import necessary libraries. First of all preprocess your data, then build a model, train your dataset finally you can do prediction part

Upvotes: 0

Shagun Sodhani
Shagun Sodhani

Reputation: 3727

I understand that the fundamental problem that you are facing is "how to train RNNs (LSTM is a kind of RNN afterall) using Keras". I would point you to this excellent collection of sample codes in Keras Github repository.

This is a simple script showing how to train LSTMs. You should be able to run this script as is. To answer why you are getting that specific error undefined name 'SimpleRNN', it seems you forgot to import SimpleRNN. Try following the script/link I shared and let me know if it works for you :)

Upvotes: 2

Related Questions