Closed jabowery closed 4 years ago
What do you mean by "need to flatten the signal features
" exactly? Can you give an example?
I figured out how to do what I wanted by strategic placement of the flatten()
call. I was taking the signal_features
, flatten()
ing them into a single state vector and feeding them into a pytorch
Linear
layer in parallel, rather than treating them as a time series of features. So the number of dimensions in the state space is a factor of window_size
larger than normal. Where the "higher order" matrix multiplications come in is an optimization I was doing to use the same feed forward call I used on single instances to do batch
instances all at once in the same call by adding a dimension.
I'm finding it difficult to use this because of the need to flatten the
signal_features
into a single vector so as to simplify later shapes for higher order matrix multiplications.