Research map / Optimization / Nonconvex and smooth optimization
Generative Adversarial Imitation Learning with Neural Network Parameterization: Global Optimality and Convergence Rate
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 |
|---|---|---|---|
| Regularizer lambda (psi, convex L-Lipschitz) added to GAIL objective with regularization parameter lambda | GAIL with two-layer neural networks · theoretical bound (Theorem 4.5) | expected R-distance optimality gap | No clear effect |
| Step size eta set to 1/sqrt(T) and batch size N grown with T (N = Omega(B^2 T sigma^2)) | GAIL alternating natural policy gradient + gradient ascent with neural networks · theoretical convergence rate | expected R-distance optimality gap | Helped |
| Overparameterized width m increased (m = Omega(B^10 T^6)) and TD iterations T_TD = Omega(m) in neural TD critic | Neural TD critic in GAIL · theoretical error bound (Proposition 4.4) | Q-function estimation error | Helped |
| Random initialization symmetric variant (W0 duplicated with sign-flipped b) so u0 = 0 | Neural network initialization in GAIL · theoretical analysis | approximation error term | Helped |
Closest papers by meaning
- On Computation and Generalization of Generative Adversarial Imitation Learning (ICLR 2020 · similarity 0.96)
- C-GAIL: Stabilizing Generative Adversarial Imitation Learning with Control Theory (NeurIPS 2024 · similarity 0.94)
- Exploring Gradient Explosion in Generative Adversarial Imitation Learning: A Probabilistic Perspective (AAAI 2024 · similarity 0.94)
- Provably and Practically Efficient Adversarial Imitation Learning with General Function Approximation (NeurIPS 2024 · similarity 0.89)
- Diffusion-Reward Adversarial Imitation Learning (NeurIPS 2024 · similarity 0.89)
- Diffusing States and Matching Scores: A New Framework for Imitation Learning (ICLR 2025 · 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.