Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-8423
Main Title: Optimizing the depth and the direction of prospective planning using information values
Author(s): Sezener, Can Eren
Dezfouli, Amir
Keramati, Mehdi
Type: Article
Language Code: en
Abstract: Evaluating the future consequences of actions is achievable by simulating a mental search tree into the future. Expanding deep trees, however, is computationally taxing. Therefore, machines and humans use a plan-until-habit scheme that simulates the environment up to a limited depth and then exploits habitual values as proxies for consequences that may arise in the future. Two outstanding questions in this scheme are “in which directions the search tree should be expanded?”, and “when should the expansion stop?”. Here we propose a principled solution to these questions based on a speed/accuracy tradeoff: deeper expansion in the appropriate directions leads to more accurate planning, but at the cost of slower decision-making. Our simulation results show how this algorithm expands the search tree effectively and efficiently in a grid-world environment. We further show that our algorithm can explain several behavioral patterns in animals and humans, namely the effect of time-pressure on the depth of planning, the effect of reward magnitudes on the direction of planning, and the gradual shift from goal-directed to habitual behavior over the course of training. The algorithm also provides several predictions testable in animal/human experiments.
URI: https://depositonce.tu-berlin.de/handle/11303/9365
http://dx.doi.org/10.14279/depositonce-8423
Issue Date: 12-Mar-2019
Date Available: 29-Apr-2019
DDC Class: 570 Biowissenschaften; Biologie
004 Datenverarbeitung; Informatik
610 Medizin und Gesundheit
Subject(s): search tree
Monte Carlo method
decision making
behavioral pattern
Sponsor/Funder: DFG, 414044773, Open Access Publizieren 2019 - 2020 / Technische Universität Berlin
License: https://creativecommons.org/licenses/by/4.0/
Journal Title: PLOS Computational Biology
Publisher: Public Library of Science
Publisher Place: San Francisco, Calif.
Volume: 15
Issue: 3
Article Number: e1006827
Publisher DOI: 10.1371/journal.pcbi.1006827
EISSN: 1553-7358
ISSN: 1553-734X
Appears in Collections:FG Neuronale Informationsverarbeitung » Publications

Files in This Item:
File Description SizeFormat 
journal.pcbi.1006827.pdf1.72 MBAdobe PDFView/Open


This item is licensed under a Creative Commons License Creative Commons