Modeling goal selection with program synthesis

J. Branson Byers; Bonan Zhao; Yael Niv

Modeling goal selection with program synthesis

J. Branson Byers, Bonan Zhao, Yael Niv

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

Abstract

In reinforcement learning, it can be difficult to select goals among many possible states. We define a framework for understanding optimal goal selection and its computational cost. We then propose program induction as a method for defining human-like priors that make informed goal selection easier. By generating programs that map to a state space and reward function, we efficiently approximate an optimal goal selecting agent. We highlight applications of this work to sequential goal selection and modeling of human behavior.

Original language	English
Pages	1-7
Number of pages	7
Publication status	Published - 9 Oct 2024
Externally published	Yes
Event	The 6th International Workshop on Intrinsically Motivated Open-ended Learning - Vancouver Convention Center, Vancouver, Canada Duration: 15 Dec 2024 → 15 Dec 2024 Conference number: 6 https://imol-workshop.github.io/

Workshop

Workshop	The 6th International Workshop on Intrinsically Motivated Open-ended Learning
Abbreviated title	IMOL 2024
Country/Territory	Canada
City	Vancouver
Period	15/12/24 → 15/12/24
Internet address	https://imol-workshop.github.io/

Keywords / Materials (for Non-textual outputs)

reinforcement learning
program inductions
goals
autonomous agents

Access to Document

https://openreview.net/forum?id=DfFE7hfnEbLicence: Creative Commons: Attribution (CC-BY)

Cite this

@conference{f2f9f283e70948e2ae564662487d3e70,

title = "Modeling goal selection with program synthesis",

abstract = "In reinforcement learning, it can be difficult to select goals among many possible states. We define a framework for understanding optimal goal selection and its computational cost. We then propose program induction as a method for defining human-like priors that make informed goal selection easier. By generating programs that map to a state space and reward function, we efficiently approximate an optimal goal selecting agent. We highlight applications of this work to sequential goal selection and modeling of human behavior.",

keywords = "reinforcement learning, program inductions, goals, autonomous agents",

author = "Byers, \{J. Branson\} and Bonan Zhao and Yael Niv",

year = "2024",

month = oct,

day = "9",

language = "English",

pages = "1--7",

note = "The 6th International Workshop on Intrinsically Motivated Open-ended Learning, IMOL 2024 ; Conference date: 15-12-2024 Through 15-12-2024",

url = "https://imol-workshop.github.io/",

}

TY - CONF

T1 - Modeling goal selection with program synthesis

AU - Byers, J. Branson

AU - Zhao, Bonan

AU - Niv, Yael

N1 - Conference code: 6

PY - 2024/10/9

Y1 - 2024/10/9

N2 - In reinforcement learning, it can be difficult to select goals among many possible states. We define a framework for understanding optimal goal selection and its computational cost. We then propose program induction as a method for defining human-like priors that make informed goal selection easier. By generating programs that map to a state space and reward function, we efficiently approximate an optimal goal selecting agent. We highlight applications of this work to sequential goal selection and modeling of human behavior.

AB - In reinforcement learning, it can be difficult to select goals among many possible states. We define a framework for understanding optimal goal selection and its computational cost. We then propose program induction as a method for defining human-like priors that make informed goal selection easier. By generating programs that map to a state space and reward function, we efficiently approximate an optimal goal selecting agent. We highlight applications of this work to sequential goal selection and modeling of human behavior.

KW - reinforcement learning

KW - program inductions

KW - goals

KW - autonomous agents

M3 - Poster

SP - 1

EP - 7

T2 - The 6th International Workshop on Intrinsically Motivated Open-ended Learning

Y2 - 15 December 2024 through 15 December 2024

ER -

Modeling goal selection with program synthesis

Abstract

Workshop

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this