True loss and true variance

[danil-z]

Hi ,

Thanks for great library!

Could you support minimize function based true_loss and true_loss_variance ?

https://github.com/hyperopt/hyperopt/wiki/FMin

I mean if model returns dictionary containing these keys , then it's probably better select model based on generalization loss ?!

[maxpumperla]

hyperopt does this for you implicitly when you pass these extra parameters, would be in this line:

https://github.com/maxpumperla/hyperas/blob/master/hyperas/optim.py#L73

to use this in hyperas, just pass them here:

https://github.com/maxpumperla/hyperas/blob/master/examples/simple.py#L65