Research map / Optimization / Optimizers
On the Limitations of First-Order Approximation in GAN Dynamics
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 optimal discriminator updates instead of simultaneous first-order discriminator updates | GMM-GAN · Two-Gaussian mixture density estimation | TV distance: d_{[0mTV} (G_{[0mμ^*}, G_{[0mμ^(T)}) ≤ δ | Helped |
| Use simultaneous first-order updates for both generator and discriminator | GMM-GAN · Random initializations | convergence to truth | Hurt |
| Unroll 5 discriminator gradient steps | GMM-GAN · Random initializations | convergence to truth | Mixed |
| Use the alternative absolute-value formulation of total variation | First-order GMM-GAN dynamics · Random discriminator initializations | convergence to truth | Mixed |
| Use the alternative absolute-value formulation of total variation | Unrolled GMM-GAN dynamics · Random discriminator initializations | convergence to truth | Mixed |
| Initialize both generator means to the same value | GMM-GAN dynamics · Two-Gaussian mixture density estimation | convergence to truth | Hurt |
Closest papers by meaning
- A Neural Tangent Kernel Perspective of GANs (ICML 2022 · similarity 0.92)
- Understanding and Stabilizing GANs’ Training Dynamics Using Control Theory (ICML 2020 · similarity 0.91)
- Smoothness and Stability in GANs (ICLR 2020 · similarity 0.91)
- Approximation and Convergence Properties of Generative Adversarial Learning (NeurIPS 2017 · similarity 0.91)
- Instability and Local Minima in GAN Training with Kernel Discriminators (NeurIPS 2022 · similarity 0.91)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.90)
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.