HNGlance

Top Best New Show

Idempotent generative network – faster alternative to GAN

https://assafshocher.github.io/IGN/

by ToJans • 3 years ago

2 1 3 years ago

AS

assafshocher.github.io

Related Stories

Loading related stories...

Comments

1

yorwba• 3 years ago

I was curious how they avoid trivial idempotent functions like f(x) = x and it turns out that when they apply the network twice to an image of random noise to check for idempotence, the first instance of the network is rewarded for producing something that isn't changed much by the second instance (encouraging idempotence) but the second instance is punished for not changing an image that was produced by the first instance instead of being sampled from the ground truth (discouraging idempotence).

This technique has a GAN-like adversarial flavor to it, complete with stability issues due to competing objectives. Still, amazing that it works at all.

Source preview

assafshocher.github.io

HNGlance

© 2024-2026 HN Glance. Story content belongs to its original sources.

Privacy
Terms