Sentiment Analysis: Using Recurrent Neural Networks

https://d2l.ai/chapter_natural-language-processing-applications/sentiment-analysis-rnn.html

Why don’t we need num_steps here? Recall section 9.2 LSTM needs it.

num_steps is defined when preprocessing the dataset:

https://d2l.ai/chapter_natural-language-processing-applications/sentiment-analysis-and-dataset.html

(batch size, 4 * no. of hidden units), why β€œ4” here in fc output?

1 Like

I guess 3 is the right answer: