Hi everybody, I am looking for some documentation regarding how to improve a model built from scratch. Like should I add more layers, more neurons, how to interpret the loss evolution and so on. If you have good resources on that topic that would be very much appreciated. Thanks
Unfortunately there is no overall straightforward method for choosing your layers, neurons etc. You could try looking at other implementations for the same idea, but it mostly comes to trial and error: Testing your model, evaluating and tweaking your hyperparameters and network structure
1 Like
Thanks for your reply.
Ok, “try and fail” then
1 Like