Poster Presentation: Learning to Plan with Tree Search via Deep RL

Dylan Cope

Poster Presentation: Learning to Plan with Tree Search via Deep RL

Informatics

Research output: Contribution to conference types › Poster › peer-review

Abstract

Tree search is an important component of many decision-making algorithms but often relies on an evaluation function that estimates the desirability of each node. In this paper, we propose to learn which nodes to expand based on a variety of object-level features. We introduce a reward function for this problem based on value of computation estimates with respect to improving the policy for the underlying problem. We apply deep reinforcement learning to this problem in an approach we call Reinforcement Learning for Tree Search (RLTS) and demonstrate that it can yield better performance than baselines in a procedurally generated environment.

Original language	English
Publication status	Accepted/In press - 20 Aug 2023
Event	The Bridging the Gap Between AI Planning and Reinforcement Learning Workshop at the International Joint Conference on AI - Macau, China Duration: 20 Aug 2023 → 20 Aug 2023 https://prl-theworkshop.github.io/prl2023-ijcai/

Workshop

Workshop	The Bridging the Gap Between AI Planning and Reinforcement Learning Workshop at the International Joint Conference on AI
Abbreviated title	PRL @ IJCAI 2023
Country/Territory	China
City	Macau
Period	20/08/2023 → 20/08/2023
Internet address	https://prl-theworkshop.github.io/prl2023-ijcai/

Keywords

Reinforcement Learning
Planning Algorithms
Tree Search
Metareasoning

Cite this

@conference{7bbf1bb50f7542dda12d4eccf7428600,

title = "Poster Presentation: Learning to Plan with Tree Search via Deep RL",

abstract = "Tree search is an important component of many decision-making algorithms but often relies on an evaluation function that estimates the desirability of each node. In this paper, we propose to learn which nodes to expand based on a variety of object-level features. We introduce a reward function for this problem based on value of computation estimates with respect to improving the policy for the underlying problem. We apply deep reinforcement learning to this problem in an approach we call Reinforcement Learning for Tree Search (RLTS) and demonstrate that it can yield better performance than baselines in a procedurally generated environment.",

keywords = "Reinforcement Learning, Planning Algorithms, Tree Search, Metareasoning",

author = "Dylan Cope",

year = "2023",

month = aug,

day = "20",

language = "English",

note = "The Bridging the Gap Between AI Planning and Reinforcement Learning Workshop at the International Joint Conference on AI , PRL @ IJCAI 2023 ; Conference date: 20-08-2023 Through 20-08-2023",

url = "https://prl-theworkshop.github.io/prl2023-ijcai/",

}

TY - CONF

T1 - Poster Presentation: Learning to Plan with Tree Search via Deep RL

AU - Cope, Dylan

PY - 2023/8/20

Y1 - 2023/8/20

N2 - Tree search is an important component of many decision-making algorithms but often relies on an evaluation function that estimates the desirability of each node. In this paper, we propose to learn which nodes to expand based on a variety of object-level features. We introduce a reward function for this problem based on value of computation estimates with respect to improving the policy for the underlying problem. We apply deep reinforcement learning to this problem in an approach we call Reinforcement Learning for Tree Search (RLTS) and demonstrate that it can yield better performance than baselines in a procedurally generated environment.

AB - Tree search is an important component of many decision-making algorithms but often relies on an evaluation function that estimates the desirability of each node. In this paper, we propose to learn which nodes to expand based on a variety of object-level features. We introduce a reward function for this problem based on value of computation estimates with respect to improving the policy for the underlying problem. We apply deep reinforcement learning to this problem in an approach we call Reinforcement Learning for Tree Search (RLTS) and demonstrate that it can yield better performance than baselines in a procedurally generated environment.

KW - Reinforcement Learning

KW - Planning Algorithms

KW - Tree Search

KW - Metareasoning

M3 - Poster

T2 - The Bridging the Gap Between AI Planning and Reinforcement Learning Workshop at the International Joint Conference on AI

Y2 - 20 August 2023 through 20 August 2023

ER -

Poster Presentation: Learning to Plan with Tree Search via Deep RL

Abstract

Workshop

Keywords

Fingerprint

Cite this