Active observing in continuous-time control

S Holt, A Hüyük… - Advances in Neural …, 2024 - proceedings.neurips.cc
The control of continuous-time environments while actively deciding when to take costly
observations in time is a crucial yet unexplored problem, particularly relevant to real-world …

An Idiosyncrasy of Time-discretization in Reinforcement Learning

K De Asis, RS Sutton - arxiv preprint arxiv:2406.14951, 2024 - arxiv.org
Many reinforcement learning algorithms are built on an assumption that an agent interacts
with an environment over fixed-duration, discrete time steps. However, physical systems are …

Mitigating the curse of horizon in Monte-Carlo returns

A Ayoub, D Szepesvari, F Zanini, B Chan… - Reinforcement …, 2024 - openreview.net
The standard framework in reinforcement learning (RL) dictates that an agent should use
every observation collected from interactions with the environment when updating its value …

Reinforcement Learning Through the Lens of Koopman Operators-The Infinite-Dimensional Framework

F Zanini - 2023 - research.unipd.it
Reinforcement Learning is nowadays one of the most active research areas in Artificial
Intelligence. This is due both to its practical successes, and to the theoretical challenges it …