Research map / Image generation / GANs
Composite Functional Gradient Learning of Generative Adversarial Models
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 |
|---|---|---|---|
| Use xICFG instead of the strongest baseline | Image generation with convolutional networks · MNIST | inception score: 9.57 -> 9.78 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with fully-connected generators · MNIST | inception score: 9.45 -> 9.72 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with convolutional networks · SVHN | inception score: 8.33 -> 8.72 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with fully-connected generators · SVHN | inception score: 7.81 -> 8.16 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with ResNets · LSUN BR+LR | inception score: 1.56 -> 1.64 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with fully-connected generators · LSUN BR+LR | inception score: 1.39 -> 1.50 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with convolutional networks · LSUN T+B | inception score: 1.76 -> 1.80 | Helped |
| Use xICFG instead of the strongest baseline | Image generation with fully-connected generators · LSUN T+B | inception score: 1.68 -> 1.74 | Helped |
| Reduce xICFG iteration count T from 25 to 5 on SVHN, 10 on MNIST, or 15 on LSUN (ablation) | xICFG with convolutional approximators · MNIST, SVHN, and LSUN image generation | inception score | No clear effect |
| Use the final approximator as the generator instead of the final ICFG generator (ablation) | xICFG · Image generation | inception score | Hurt |
| Train the discriminator to convergence at every functional-gradient iteration (ablation) | CFG · Image generation | generator quality | Hurt |
| Set T to 1 and use only one approximator gradient step on one mini-batch (ablation) | xICFG, yielding the GAN-style update · Image generation | training stability | Hurt |
Closest papers by meaning
- TRAINING GENERATIVE ADVERSARIAL NETWORKS VIA PRIMAL-DUAL SUBGRADIENT METHODS: A LAGRANGIAN PERSPECTIVE ON GAN (ICLR 2018 · similarity 0.90)
- On Stabilizing Generative Adversarial Training With Noise (CVPR 2019 · similarity 0.89)
- Unrolled Generative Adversarial Networks (ICLR 2017 · similarity 0.89)
- Stabilizing Training of Generative Adversarial Networks through Regularization (NeurIPS 2017 · similarity 0.89)
- Towards Principled Methods for Training Generative Adversarial Networks (ICLR 2017 · similarity 0.89)
- AdaGAN: Boosting Generative Models (NeurIPS 2017 · similarity 0.89)
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.