0
$\begingroup$

In regards to this paper, arXiv:1611.01578 [cs.LG] the very last image, seen below, has a graph of a LSTM Cell (top left). I am trying to recreate it in python, but main thing that throwing me off.

Why are input and hidden state added in this diagram? Everywhere I see an LSTM diagram, the input and hidden state are concatenated, multiplied with a weight, and then bias is added. This matrix is then split into gates, and Each gate undergoes operations and you end up with hidden state and cell state. Last Page of NAS paper

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

It is simply the dot product rewritten as addition where you add the multiplication of rows by columns which equals the dot product of the concatenation.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.