Reinforcement Learning
DeepTOP: Deep Threshold-Optimal Policy for MDPs and RMABs
We consider the problem of learning the optimal threshold policy for control problems.
Learning Active Camera for Multi-Object Navigation Peihao Chen 1,6 Dongyu Ji1 Kunyang Lin
As a result, the agent may take a very long time to perceive the environment due to limited camera scope. In contrast, humans typically gain a larger field of view by looking around for a better perception of the environment.
Regret Bounds for Information-Directed Reinforcement Learning
As a result, IDS demonstrates impressive empirical performance [Russo and V an Roy, 2018] and outperforms UCB and TS in terms of asymptotic optimality [Kirschner et al.,
Assistive Teaching of Motor Control Tasks to Humans
However, these approaches often fail to account for humans' ability to
No RL, No Simulation: Learning to Navigate without Navigating
Our approach, No RL, No Simulator (NRNS), is simple and scalable, yet highly effective. NRNS outperforms RL-based formulations by a significant margin.