I manipulate the time series using the different structures of the neural networks in order to make a prediction, and I wonder if there is a way to choose the parameters of the networks intelligently? from the characteristics of the signal, namely (trend, seasonality ...) can we choose these parameters that will make learning better?
$\begingroup$
$\endgroup$
2
-
$\begingroup$ Many folks use grid search. Some crazy folks use bounded Nelder Mead. $\endgroup$– EngrStudentAug 14, 2020 at 14:14
-
$\begingroup$ Can you please be more specific on what you have already tried and what is blocking you from doing what you are aiming for. $\endgroup$– hH1sG0n3May 4, 2021 at 10:33
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Indeed you can introduce some "unvariant" features to your LSTM network using Conditional RNN that use these features to create the initial hidden state:
https://github.com/philipperemy/cond_rnn
I hope this is what you are looking for.
