Research map / Label-efficient and robust learning / Semi-supervised learning
Improving GAN Training via Binarized Representation Entropy (BRE) Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add BRE regularization to feature matching | Semi-supervised GAN · CIFAR10, 1000 labeled examples | test error rate (%): 17.42 \pm 0.50 -> 16.98 \pm 0.52 | Helped |
| Add BRE regularization to feature matching | Semi-supervised GAN · CIFAR10, 1000 labeled examples | train classification loss: \expnumber{9.25}{-4} \pm \expnumber{5.05}{-4} -> \expnumber{5.03}{-4} \pm \expnumber{3.50}{-4} | Helped |
| Add BRE regularization to feature matching | 10-model ensemble · CIFAR10, 1000 labeled examples | test error rate (%): 14.25 -> 13.93 | Helped |
| Add discriminator reconstruction loss (ablation) | GAN discriminator with layer normalization · GAN auxiliary reconstruction experiment | Inception score: 6.4025 \pm 0.2187 -> 6.1958 \pm 0.2438 (lambda_recon = 10) | Hurt |
| Add discriminator reconstruction loss (ablation) | GAN discriminator with batch normalization · GAN auxiliary reconstruction experiment | Inception score: 6.5865 \pm 0.1837 -> 6.5356 \pm 0.2176 (lambda_recon = 1) | Hurt |
| Use tanh instead of scale-adaptive softsign in the BRE approximation (ablation) | BRE-regularized GAN · CIFAR10 GAN training | Inception score | Hurt |
| Apply BRE to discriminator hidden layers | GAN · 2D mixture of Gaussians | mode coverage | Helped |
| Apply BRE to discriminator hidden layers | DCGAN · CIFAR10 | Inception score convergence speed | Helped |
| Apply BRE to discriminator hidden layers | DCGAN with aggressive optimization · CIFAR10 | final Inception score | Helped |
| Apply BRE to discriminator hidden layers | Semi-supervised GAN · SVHN | training stability | Helped |
Closest papers by meaning
- Stabilizing Training of Generative Adversarial Networks through Regularization (NeurIPS 2017 · similarity 0.90)
- Mode Regularized Generative Adversarial Networks (ICLR 2017 · similarity 0.89)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.88)
- Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · similarity 0.88)
- BinGAN: Learning Compact Binary Descriptors with a Regularized GAN (NeurIPS 2018 · similarity 0.88)
- MS$^3$D: A RG Flow-Based Regularization for GAN Training with Limited Data (ICML 2024 · similarity 0.87)
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.