Why is the discriminator necessary in a GAN? - Deep Learning: Getting Started - DeepTalk - Deep Learning Community

Why is the discriminator necessary in a GAN?

post by cgold on Apr 5, 2020

I am currently attempting to code a GAN to learn to generate images of hand drawn digits (MNIST dataset) and am having little success. My discriminator / generator are failing to converge and my generator output is noise.

After trying to solve the problem for a while, I tried not using the discriminator at all, and just training the generator with a desired output of a sampled image, and within minutes, the generator had learned to output very realistic images.

Why is this a bad idea? I haven’t seen any use of this method online and it is not used in GANs, however seems to be much simpler and works (for me) much better. Does it overfit too much or is there another problem?

Thanks

1.6k views

Related topics

Topic	Replies	Views	Activity
About Deep Learning General (DLG) Deep Learning: General Discussion	0	2.1k	Oct 2017
Project approach consultation - RF / DL / …? Deep Learning: General Discussion	0	1.5k	Oct 2019
Are gamma and beta just for removing extra 1/(epsilon)^(½) when mu and sigma all equal zero? I don’t get why use z tilde in Andrew’s deep learning course? Deep Learning: General Discussion	0	1.2k	Aug 2020
Basic idea of Convolution Neutral Network Technical Help	0	1.2k	Jul 2020
Deep Learning Community Forum & Discord Server Deep Learning: General Discussion	1	5.1k	Oct 2021