a4(part 2)

[ranjeetkumar]

How we feed data to RNN? In our assignment after vectorization, we have two dimension matrix as input. Suppose, we have a matrix with shape (m,n), how we will feed this input to the RNN? How to decide the batch size, no of batches, series length etc? Can anyone please explain?

[aronwc]

I'm a little confused by your question. The input is (m,n), but this is for all tokens in a sentence. Each single token has a vector representation. See, e.g., in L19, how we created a vector for each token in the train_all method.

inputs = [char_to_ix[ch] for ch in data.lower()[p:p+seq_length]]

You might also look at the ptb_producer method from the tensor flow tutorial for example: https://github.com/tensorflow/models/blob/master/tutorials/rnn/ptb/reader.py

This is used to create the input data structure for the RNN https://github.com/tensorflow/models/blob/master/tutorials/rnn/ptb/ptb_word_lm.py

-Aron

On Tue, Apr 18, 2017 at 2:13 PM, Ranjeet kumar notifications@github.com wrote:

How we feed data to RNN? In our assignment after vectorization, we have two dimension matrix as input. Suppose, we have a matrix with shape (m,n), how we will feed this input to the RNN? How to decide the batch size, no of batches, series length etc? Can anyone please explain?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/iit-cs585/assignments/issues/33, or mute the thread https://github.com/notifications/unsubscribe-auth/ADv-hXFFr1_A2GZ6mePaApkgE3VZSaXEks5rxQtmgaJpZM4NAxXT .

[karthikshivaram24]

Professor, How would you divide the dataset into matrix of shape = [Batch_size, Timesteps, InputDimensions] ?

[ranjeetkumar]

Thanks all