Search papers, labs, and topics across Lattice.
Seoul National University, Purdue University
1
0
1
Learning policies from just one trajectory in average-reward MDPs is now feasible, with guarantees that could transform how we approach sample efficiency in reinforcement learning.
