OntheConvergenceTheoryofDebiased Model-AgnosticMeta-ReinforcementLearning
–Neural Information Processing Systems
In particular, using stochastic gradients in MAML update steps is crucial for RL problems since computation of exact gradients requires access to a large number of possible trajectories.
Neural Information Processing Systems
Feb-7-2026, 16:24:40 GMT
- Country:
- Europe > Sweden
- North America > United States
- California (0.04)
- Massachusetts > Middlesex County
- Cambridge (0.04)
- Oceania > Australia
- New South Wales > Sydney (0.04)
- Genre:
- Research Report (0.46)
- Technology: