Confusion with optimizations.

torch / tutorials

A series of machine learning tutorials for Torch7

622 stars 312 forks source link

Closed KHN190 closed 9 years ago

KHN190 commented 9 years ago

I noticed in the optim methods there's a required feval to be passed through. For example:

supervised training, line 145, which calculated f(x) and df / dx. What is this f supposed to do?

Is f a loss function, then what is x? In some code (e.x. LSTM), why they weren't using optim?

soumith commented 9 years ago

f is the optimization function (for example: sgd), and x are the parameters of your neural network.