Instance Selection for Dynamic Algorithm Configuration with Reinforcement Learning: Improving Generalization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Benjamins, Carolin, Cenikj, Gjorgjina, Nikolikj, Ana, Mohan, Aditya, Eftimov, Tome, Lindauer, Marius
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917726853791744
author Benjamins, Carolin
Cenikj, Gjorgjina
Nikolikj, Ana
Mohan, Aditya
Eftimov, Tome
Lindauer, Marius
author_facet Benjamins, Carolin
Cenikj, Gjorgjina
Nikolikj, Ana
Mohan, Aditya
Eftimov, Tome
Lindauer, Marius
contents Dynamic Algorithm Configuration (DAC) addresses the challenge of dynamically setting hyperparameters of an algorithm for a diverse set of instances rather than focusing solely on individual tasks. Agents trained with Deep Reinforcement Learning (RL) offer a pathway to solve such settings. However, the limited generalization performance of these agents has significantly hindered the application in DAC. Our hypothesis is that a potential bias in the training instances limits generalization capabilities. We take a step towards mitigating this by selecting a representative subset of training instances to overcome overrepresentation and then retraining the agent on this subset to improve its generalization performance. For constructing the meta-features for the subset selection, we particularly account for the dynamic nature of the RL agent by computing time series features on trajectories of actions and rewards generated by the agent's interaction with the environment. Through empirical evaluations on the Sigmoid and CMA-ES benchmarks from the standard benchmark library for DAC, called DACBench, we discuss the potentials of our selection technique compared to training on the entire instance set. Our results highlight the efficacy of instance selection in refining DAC policies for diverse instance spaces.
format Preprint
id arxiv_https___arxiv_org_abs_2407_13513
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Instance Selection for Dynamic Algorithm Configuration with Reinforcement Learning: Improving Generalization
Benjamins, Carolin
Cenikj, Gjorgjina
Nikolikj, Ana
Mohan, Aditya
Eftimov, Tome
Lindauer, Marius
Machine Learning
Dynamic Algorithm Configuration (DAC) addresses the challenge of dynamically setting hyperparameters of an algorithm for a diverse set of instances rather than focusing solely on individual tasks. Agents trained with Deep Reinforcement Learning (RL) offer a pathway to solve such settings. However, the limited generalization performance of these agents has significantly hindered the application in DAC. Our hypothesis is that a potential bias in the training instances limits generalization capabilities. We take a step towards mitigating this by selecting a representative subset of training instances to overcome overrepresentation and then retraining the agent on this subset to improve its generalization performance. For constructing the meta-features for the subset selection, we particularly account for the dynamic nature of the RL agent by computing time series features on trajectories of actions and rewards generated by the agent's interaction with the environment. Through empirical evaluations on the Sigmoid and CMA-ES benchmarks from the standard benchmark library for DAC, called DACBench, we discuss the potentials of our selection technique compared to training on the entire instance set. Our results highlight the efficacy of instance selection in refining DAC policies for diverse instance spaces.
title Instance Selection for Dynamic Algorithm Configuration with Reinforcement Learning: Improving Generalization
topic Machine Learning
url https://arxiv.org/abs/2407.13513