AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Off-Policy Reinforcement Learning for Efficient and Effective GAN Architecture Search

Yuan Tian, Qin Wang, Zhiwu Huang, Wen Li, Dengxin Dai, Minghao Yang, Jun Wang, Olga Fink

ECCV 2020 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 63 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add FID term (alpha=0.01) to the IS-only reward (ablation)E2GAN · CIFAR-10 unconditionalFID: 15.64 -> 11.26Improved
add FID term (alpha=0.01) to the IS-only reward (ablation)E2GAN · CIFAR-10 unconditionalInception Score: 8.81 -> 8.51Hurt
add FID term to the reward (ablation)E2GAN · CIFAR-10 unconditionalsearch cost (GPU days): 0.1 -> 0.3Hurt
replace AutoGAN on-policy multi-level search with off-policy MDP search (E2GAN)GAN architecture search · CIFAR-10 unconditionalFID: 12.42 -> 11.26Improved
replace AutoGAN with E2GAN searchGAN architecture search · CIFAR-10 unconditionalInception Score: 8.55 -> 8.51No clear effect
replace AutoGAN with E2GAN searchGAN architecture search · CIFAR-10 unconditionalsearch cost (GPU days): 2 -> 0.3Improved
replace AGAN with E2GAN searchGAN architecture search · CIFAR-10 unconditionalsearch cost (GPU days): 1200 -> 0.3Improved
use E2GAN-top2 instead of AutoGAN-top2GAN architecture search · CIFAR-10 unconditionalFID: 13.67 -> 12.96Improved
transfer E2GAN architecture instead of AutoGAN architecture to STL-10GAN architecture search · STL-10 unconditionalFID: 31.01 -> 25.35Improved
transfer E2GAN architecture instead of AutoGAN architecture to STL-10GAN architecture search · STL-10 unconditionalInception Score: 9.16 -> 9.51Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy