GAN Feature Matching

Introduced by Salimans et al. in Improved Techniques for Training GANs

Feature Matching is a regularizing objective for a generator in generative adversarial networks that prevents it from overtraining on the current discriminator. Instead of directly maximizing the output of the discriminator, the new objective requires the generator to generate data that matches the statistics of the real data, where we use the discriminator only to specify the statistics that we think are worth matching. Specifically, we train the generator to match the expected value of the features on an intermediate layer of the discriminator. This is a natural choice of statistics for the generator to match, since by training the discriminator we ask it to find those features that are most discriminative of real data versus data generated by the current model.

Letting $\mathbf{f}\left(\mathbf{x}\right)$ denote activations on an intermediate layer of the discriminator, our new objective for the generator is defined as: $ ||\mathbb{E}_{x\sim p_{data} } \mathbf{f}\left(\mathbf{x}\right) − \mathbb{E}_{\mathbf{z}∼p_{\mathbf{z}}\left(\mathbf{z}\right)}\mathbf{f}\left(G\left(\mathbf{z}\right)\right)||^{2}_{2} $. The discriminator, and hence $\mathbf{f}\left(\mathbf{x}\right)$, are trained as with vanilla GANs. As with regular GAN training, the objective has a fixed point where G exactly matches the distribution of training data.

Source: Improved Techniques for Training GANs

Latest Papers

PAPER DATE
Imbalanced Data Learning by Minority Class Augmentation using Capsule Adversarial Networks
Pourya ShamsolmoaliMasoumeh ZareapoorLinlin ShenAbdul Hamid SadkaJie Yang
2020-04-05
Feature Quantization Improves GAN Training
| Yang ZhaoChunyuan LiPing YuJianfeng GaoChangyou Chen
2020-04-05
Learning the Loss Functions in a Discriminative Space for Video Restoration
Younghyun JoJaeyeon KangSeoung Wug OhSeonghyeon NamPeter VajdaSeon Joo Kim
2020-03-20
Image Fine-grained Inpainting
| Zheng HuiJie LiXiumei WangXinbo Gao
2020-02-07
Semi-Supervised Self-Growing Generative Adversarial Networks for Image Recognition
Haoqian WangZhiwei XuJun XuWangpeng AnLei ZhangQionghai Dai
2019-08-11
Few-shot 3D Multi-modal Medical Image Segmentation using Generative Adversarial Learning
| Arnab Kumar MondalJose DolzChristian Desrosiers
2018-10-29
A Generative Model of Textures Using Hierarchical Probabilistic Principal Component Analysis
Aiga SuzukiHayaru Shouno
2018-10-16
Out-of-domain Detection based on Generative Adversarial Network
Seonghan RyuSangjun KooHwanjo YuGary Geunbae Lee
2018-10-01
Adversarial Training for Adverse Conditions: Robust Metric Localisation using Appearance Transfer
Horia PoravWill MaddernPaul Newman
2018-03-09
Novelty Detection with GAN
Mark KligerShachar Fleishman
2018-02-28
Bayesian GAN
| Yunus SaatciAndrew G. Wilson
2017-12-01
Bayesian GAN
| Yunus SaatchiAndrew Gordon Wilson
2017-05-26
CVAE-GAN: Fine-Grained Image Generation through Asymmetric Training
| Jianmin BaoDong ChenFang WenHouqiang LiGang Hua
2017-03-29
Improved Techniques for Training GANs
| Tim SalimansIan GoodfellowWojciech ZarembaVicki CheungAlec RadfordXi Chen
2016-06-10

Components

COMPONENT TYPE
🤖 No Components Found You can add them if they exist; e.g. Mask R-CNN uses RoIAlign

Categories