LSTM / RNN many to one

Question

I have the following dataset for a chemical process consisting of 5 consecutive input vectors to get 1 output. Each input is sampled every minute, while the output sample is selected every 5.

Although I believe that the result depends on the 5 previous input vectors, I decided to look for LSTM for my design. After much research on how my LSTM architecture should be, I came to the conclusion that I should mask part of the output sequence with zeros and leave only the last output. The final architecture is below according to my dataset:

My question is: what should be the parameters of the tensor of the three-dimensional input signal? For instance. [5, 5 ,?]? And also what should be my "lot size"? Is this the number of my samples?

Answer 1

, ( ). i.e , RNN/LSTM 5- . [batch_size, sequence_length, input_dimensionality], sequence_length 5 ( 1-5, 7-11, 13-17 ..), input_dimensionality 5 ( A - E). Batch_size ( ), 10 000 , 30-50 (. ).