Sequential Decision making in probabilistic models
Lead Research Organisation:
University of Cambridge
Department Name: Computer Science and Technology
Abstract
This proposal considers the problem of robust sequential decision making in non-linear environments. Reinforcement
learning has demonstrated high potential for solving complex problems in non-linear environments but has lacked
efficiency and robustness. We argue that in order to deploy reinforcement learning agents in the real world, it is essential to
develop similar efficiency and robustness properties that have been developed in control theory. We propose to leverage
the extensive control and probabilistic reasoning literature to improve RL algorithms and present two interesting research
directions. The first one considers using Sequential Monte-Carlo methods to improve planning for non-linear
environments. The second direction focuses on designing robust controllers by exploring the connections between
adversarial learning, robust control theory, and uncertainty modelling.
learning has demonstrated high potential for solving complex problems in non-linear environments but has lacked
efficiency and robustness. We argue that in order to deploy reinforcement learning agents in the real world, it is essential to
develop similar efficiency and robustness properties that have been developed in control theory. We propose to leverage
the extensive control and probabilistic reasoning literature to improve RL algorithms and present two interesting research
directions. The first one considers using Sequential Monte-Carlo methods to improve planning for non-linear
environments. The second direction focuses on designing robust controllers by exploring the connections between
adversarial learning, robust control theory, and uncertainty modelling.
Organisations
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/T517847/1 | 30/09/2020 | 29/09/2025 | |||
2744311 | Studentship | EP/T517847/1 | 30/09/2020 | 29/09/2023 | Pierre Thodoroff |