Peter Stone's Selected Publications

Classified by TopicClassified by Publication TypeSorted by DateSorted by First Author Last NameClassified by Funding Source

On Sampling Error in Batch Action-Value Prediction Algorithms

On Sampling Error in Batch Action-Value Prediction Algorithms.
Brahma S. Pavse, Josiah P. Hanna, Ishan Durugkar, and Peter Stone.
In In the Offline Reinforcement Learning Workshop at Neural Information Processing Systems (NeurIPS), December 2020., December 2020.
5-minute Video Presentation

Abstract

Estimating a policy's action-values is a fundamental aspect of reinforcement learning. In this work, we study the application of TD methods for learning action-values in an offline setting with a fixed batch of data. Motivated by recent work \citeppavse2020psectd, we observe that a fixed batch of offline data may contain two forms of distribution shift: the data may be collected from a different behavior policy than the target policy (off-policy data) and the empirical distribution of the data may differ from the sampling distribution of the data (sampling error). In this work, we focuson the second problem by analyzing the sampling error that arises due to variance in sampling from a finite-sized batch of data in the on-policy offline RL setting. We study how action-value learning algorithms suffer from this sampling error by considering their so-called certainty-equivalence estimates \citep sutton1988learning,pavse2020psectd. We prove that each algorithm uses its certainty-equivalence estimates of the policy and transition dynamics to converge to its respective fixed-point. Wethen empirically evaluate each algorithm's performance by measuring the mean-squared value error on Gridworld. Ultimately, we find that by reducing sampling error, an algorithm can produce significantly accurate action-valueestimations.

BibTeX Entry

@InProceedings{NeurIPS2020-Pavse,
author = {Brahma S. Pavse and Josiah P. Hanna and Ishan Durugkar and Peter Stone},
title = {On Sampling Error in Batch Action-Value Prediction Algorithms},
booktitle = {In the Offline Reinforcement Learning Workshop at Neural Information Processing Systems (NeurIPS), December 2020.},
location = {Remote (Virtual Conference)},
month = {December},
year = {2020},
abstract = {
Estimating a policy's action-values is a fundamental aspect of reinforcement
learning. In this work, we study the application of TD methods for learning
action-values in an offline setting with a fixed batch of data. Motivated by
recent work \citep{pavse2020psectd}, we observe that a fixed batch of offline
data may contain two forms of distribution shift: the data may be collected
from a different behavior policy than the target policy (off-policy data)
and the empirical distribution of the data may differ from the sampling
distribution of the data (sampling error). In this work, we focus
on the second problem by analyzing the sampling error that arises due to
variance in sampling from a finite-sized batch of data in the
\emph{on-policy offline} RL setting. We study how action-value learning
algorithms suffer from this \emph{sampling error} by considering their so-called
\emph{certainty-equivalence estimates} \citep{ sutton1988learning,pavse2020psectd}.
We prove that each algorithm uses its certainty-equivalence estimates of the
policy and transition dynamics to converge to its respective fixed-point. We
then empirically evaluate each algorithm's performance by measuring the
mean-squared value error on Gridworld. Ultimately, we find that by reducing
sampling error, an algorithm can produce significantly accurate action-value
estimations.
},
}


Generated by bib2html.pl (written by Patrick Riley ) on Thu Dec 09, 2021 19:29:32