3 ms·
> Check out "General Adversarial Networks" Another way to describe GANs is that the loss function, which used to be hand-made, now is a whole separate neural n
by visarga 10y ago
> Check out "General Adversarial Networks"
Another way to describe GANs is that the loss function, which used to be hand-made, now is a whole separate neural net and can be learned bottom-up instead of dictated top-down.
There was a time when features were hand-made. Now we have deep learning and finding good features is automatic. GANs automate another part of the process.