The authors develop a framework for last-iterate convergence of regularised policy-gradient primal-dual methods in constrained MDPs (CMDPs) with structured state spaces.
Last-iterate convergence for online constrained RL with function approximation
Authors develop a general framework achieving last-iterate guarantees for primal-dual methods in structured CMDPs without tabular assumptions.
Research Lab
Nam Phuong Tran · Trinh Ha Mai Huynh · Tuyen Pham Le · Van-Truong Nguyen · Quan Nguyen · Long Tran-Thanh
Inria centre at the University of Lille · Vin Motion · University of Warwick
Research Digest··2 min read
The authors present a general framework for last-iterate convergence in online constrained reinforcement learning with structured function approximation.
Why this paper
From Inria centre at the University of Lille and 2 others
In one line
A regularised primal-dual method achieves last-iterate convergence in structured constrained MDPs with model-free learning and improved sample complexity.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§