Last-iterate convergence on CMDPs
2Sample ComplexityDual Descent
Evaluation Results
| Method | Links | |
|---|---|---|
| Dual DescentParameterization=Softmax2024.08 | 2 | |
| PDR-ANPGParameterization=General with ϵbias > 02024.08 | 2 | |
| Lower BoundParameterization=-2024.08 | 2 | |
| Cutting-PlaneParameterization=Softmax2024.08 | 4 | |
| PDR-ANPGParameterization=General with ϵbias = 02024.08 | 4 | |
| RPG-PDParameterization=Log-linear with ϵbias = O(ϵ8)2024.08 | 6 | |
| C-PGParameterization=General2024.08 | 7 |