HF trainer's cost set wrong value.

[reyoung]

The code here set hf trainer to all monitor value of network, which include accuracy. The hf trainer will never find a new best value because the cost, which include accuracy, will always get larger if the network is trained well.

The accuracy should be not passed to hf trainer.

[lmjohns3]

I believe the HF trainer only uses the first quantity in the monitors list to evaluate training progress; the rest are just there to provide more information during training. (See https://github.com/boulanni/theano-hf/blob/master/hf.py#L268 for the code.)

[reyoung]

You're right, my mistake.