Refactor all code and think over the API (and this issue similar like #3 )
Differencing by layer maybe good idea. What about implement both variants (old, with configuration list and new with layers). They differ only in the way that weights matrices initialization occurs.
You can use inheritance and implement Abstract class of neural network with intarfaces for learning and prediction.
Soarex16
commented
6 years ago
Refactor all code and think over the API (and this issue similar like #3 )
Differencing by layer maybe good idea. What about implement both variants (old, with configuration list and new with layers). They differ only in the way that weights matrices initialization occurs.
You can use inheritance and implement Abstract class of neural network with intarfaces for learning and prediction.