Maybe useful in this context, also a little bit too late: A novel way to measure similarities between notes
Taken from here:
The use of the twelfth root of two in music is very well known in Western ...
I´m pursuing a PhD in this area so far I can tell that:
Industry is mostly based on robust yet simple models:
AWS offers a solution that as far as I can tell uses ARMA based models (AWS Forecast)
Total number of weights in LSTM N/W = 4 x inp_dim x (inp_dim + out_dim + 1)
So, in your first model:
For Stage-1(input --> h1):
inp_dim = 39; out_dim = 1024
Therefore, weights of stage-1 = 4 x 39 x ...