4 ms·
I gave it a few minutes and got nothing. Maybe try for something even simpler than MNIST, so we can get the great immediate feedback effect of http://playgroun
by SonOfLilit 9y ago
I gave it a few minutes and got nothing.
Maybe try for something even simpler than MNIST, so we can get the great immediate feedback effect of http://playground.tensorflow.org http://playground.tensorflow.org, which I consider the important aspect for learnability?
- reiinakano 9y agoDefault parameters on MNIST should give good results in 5 minutes give or take - depending on your hardware. Can you show what you got?
- EvgeniyZh 9y agoThat's what I got after 10 minutes on GTX1080 - https://imgur.com/a/4Ba9u https://imgur.com/a/4Ba9u
- reiinakano 9y agoLooks about right. The default setting is as basic as you can get. No transpose convolutions, batch norm, and other tricks for better GANs.
- DiThi 9y agoDid you click "train"? Otherwise it's just doing inference from random data.
- SonOfLilit 9y agoThanks. Doh. In Israel we have a classic song that goes "if you're cooking spaghetti and the water doesn't boil, how about turning the stove on, because that's what everybody does"
- sp332 9y agoThere's just no point getting fancy for MNIST. Using nothing but fully-connected layers and rmsprop, I got good results in less than one minute. https://imgur.com/a/7xzmQ https://imgur.com/a/7xzmQ Hidden units for discriminator layers - 100, 40, 2. Generator layers - 40, 100, 768, then "reshape" into 28x28x1.
- EvgeniyZh 9y agowhat you're showing is basically perfect discriminator with awful generator, isn't it?
- reiinakano 9y agoUnlike regular classification, it's hard to quantify the performance of a GAN by just looking at the loss curves. You really have to make a subjective evaluation by looking at the generated images. By playing around with it, the results I got from fully connected were nowhere near as good as the results I got from convolutional.