New Story
Anc-VI Sets New Standards in Speed for Bellman Consistency in Reinforcement Learning
by AnchoringJanuary 14th, 2025
Too Long; Didn't Read
Anc-VI accelerates Bellman consistency, achieving faster convergence rates for reinforcement learning, particularly when the discount factor is close to 1, outpacing standard value iteration.
Authors:
(1) Jongmin Lee, Department of Mathematical Science, Seoul National University;
(2) Ernest K. Ryu, Department of Mathematical Science, Seoul National University and Interdisciplinary Program in Artificial Intelligence, Seoul National University.
1.1 Notations and preliminaries
2.1 Accelerated rate for Bellman consistency operator
2.2 Accelerated rate for Bellman optimality opera
5 Approximate Anchored Value Iteration
6 Gauss–Seidel Anchored Value Iteration
7 Conclusion, Acknowledgments and Disclosure of Funding and References
2.1 Accelerated rate for Bellman consistency operator
First, for general state-action spaces, we present the accelerated convergence rate of Anc-VI for the Bellman consistency operator.
This paper is available on arxiv under CC BY 4.0 DEED license.
[1] Arguably, T π is affine, not linear, but we follow the convention of [69] say T π is linear.
L O A D I N G
. . . comments & more!
. . . comments & more!
