Research map / Reinforcement learning / Imitation learning
f-GAIL: Learning f-Divergence for Generative Adversarial Imitation Learning
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 |
|---|---|---|---|
| Learn the f-divergence jointly with the policy instead of fixing a standard divergence | GAIL-style imitation learning · Six control tasks | expert-normalized return | Mixed |
| Use 4 layers rather than 1 in the f* network (ablation) | f-GAIL · HalfCheetah | return: 1539$\pm$144 -> {\textbf{4445$\pm$79}} | Helped |
| Use 4 layers rather than 1 in the f* network (ablation) | f-GAIL · Reacher | return: -22.8$\pm$ 4.2 -> {\textbf{-10.6$\pm$2.6}} | Helped |
| Use 7 rather than 4 layers in the f* network (ablation) | f-GAIL · HalfCheetah | return: {\textbf{4445$\pm$79}} -> 4100 $\pm$ 51 | Hurt |
| Use 7 rather than 4 layers in the f* network (ablation) | f-GAIL · Reacher | return: {\textbf{-10.6$\pm$2.6}} -> -15.8$\pm$2.8 | Hurt |
| Increase f* network width from 25 to 100 nodes per layer (ablation) | f-GAIL · HalfCheetah | return: 3546$\pm$132 -> {\textbf{4445$\pm$79}} | Helped |
| Increase f* network width from 25 to 100 nodes per layer (ablation) | f-GAIL · Reacher | return: -25.2$\pm$ 5.35 -> {\textbf{-10.6$\pm$2.6}} | Helped |
| Increase f* network width from 100 to 200 nodes per layer (ablation) | f-GAIL · HalfCheetah | return: {\textbf{4445$\pm$79}} -> 4343 $\pm$ 80 | Hurt |
| Increase f* network width from 100 to 200 nodes per layer (ablation) | f-GAIL · Reacher | return: {\textbf{-10.6$\pm$2.6}} -> -12.6$\pm$4.0 | Hurt |
| Vary f* network depth or width (ablation) | f-GAIL · CartPole | return: 200$\pm$0 | No clear effect |
Closest papers by meaning
- Exploring Gradient Explosion in Generative Adversarial Imitation Learning: A Probabilistic Perspective (AAAI 2024 · similarity 0.92)
- On Computation and Generalization of Generative Adversarial Imitation Learning (ICLR 2020 · similarity 0.91)
- C-GAIL: Stabilizing Generative Adversarial Imitation Learning with Control Theory (NeurIPS 2024 · similarity 0.91)
- Learning from Demonstration: Provably Efficient Adversarial Policy Imitation with Linear Function Approximation (ICML 2022 · similarity 0.90)
- A Bayesian Approach to Generative Adversarial Imitation Learning (NeurIPS 2018 · similarity 0.90)
- Diffusion-Reward Adversarial Imitation Learning (NeurIPS 2024 · 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.