Keras: ValueError: Error when checking input

Question

I have pandas dataframe of length 7499042 like this:

'X'          'y'
[0.1,0.2...] 0.2
[0.3,0.4,..] 0.3
.
.

each value in pandas dataframe is numpy array of length 50. Now I am extracting it like this:

input = df['X'].values

I have layers like this:

main_input = Input(shape=(50,1), name='main_input')    
lstm_out=LSTM(32,activation='tanh',recurrent_activation='sigmoid',return_sequences=True)
mean_pooling=AveragePooling1D(pool_size=2,strides=2,padding='valid')

But when I am passing my input to this while training. It is showing the error:

ValueError: Error when checking input: expected main_input to have 3 dimensions, but got array with shape (7499042, 1)

The shape of input it is showing is (7499042,). Please help me in resolving this.

Can you please list out the shapes of the numpy array you are using as input? — Vatsal Aggarwal
– Vatsal Aggarwal, Commented Jul 3, 2018 at 13:12
@VatsalAggarwal The shape of single X is (50,) and input is (7499042,) — shivank01
– shivank01, Commented Jul 3, 2018 at 13:14

Ahmed Lahlou Mimi · Accepted Answer · 2018-07-03 13:55:37Z

1

You need to reshape your features before feeding them to the LSTM network. The LSTM layer takes a 3 dimensional input, corresponding to (batch_size, timesteps, features). This means that a single observation has to be two dimensional (timesteps, features)

In your case, a single observation is 1 dimensional (50 ,) : the whole dataset dimension is : (7499042 , 50) if the conversion was done correctly. You have to reshape your input before using it :

input = df['X'].values
input = input.reshape(input.shape[0] , input.shape[1] , 1)

in case Pandas didn't convert your initial feature to a 2d DataFrame, you have to do it before executing the above code.

answered Jul 3, 2018 at 13:55

Ahmed Lahlou Mimi

2,1413 gold badges15 silver badges21 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Keras: ValueError: Error when checking input

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related