Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Joint Entropy Search for Maximally-Informed Bayesian Optimization

Hvarfner, Carl LU ; Hutter, Frank and Nardi, Luigi LU (2022) 36th Conference on Neural Information Processing Systems, NeurIPS 2022
Abstract
Information-theoretic Bayesian optimization techniques have become popular for optimizing expensive-to-evaluate black-box functions due to their non-myopic qualities. Entropy Search and Predictive Entropy Search both consider the entropy over the optimum in the input space, while the recent Max-value Entropy Search considers the entropy over the optimal value in the output space. We propose Joint Entropy Search (JES), a novel information-theoretic acquisition function that considers an entirely new quantity, namely the entropy over the joint optimal probability density over both input and output space. To incorporate this information, we consider the reduction in entropy from conditioning on fantasized optimal input/output pairs. The... (More)
Information-theoretic Bayesian optimization techniques have become popular for optimizing expensive-to-evaluate black-box functions due to their non-myopic qualities. Entropy Search and Predictive Entropy Search both consider the entropy over the optimum in the input space, while the recent Max-value Entropy Search considers the entropy over the optimal value in the output space. We propose Joint Entropy Search (JES), a novel information-theoretic acquisition function that considers an entirely new quantity, namely the entropy over the joint optimal probability density over both input and output space. To incorporate this information, we consider the reduction in entropy from conditioning on fantasized optimal input/output pairs. The resulting approach primarily relies on standard GP machinery and removes complex approximations typically associated with information-theoretic methods. With minimal computational overhead, JES shows superior decision-making, and yields state-of-the-art performance for information-theoretic approaches across a wide suite of tasks. As a light-weight approach with superior results, JES provides a new go-to acquisition function for Bayesian optimization. (Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
in press
subject
host publication
Advances in Neural Information Processing Systems 35 (NeurIPS 2022)
publisher
Curran Associates, Inc
conference name
36th Conference on Neural Information Processing Systems, NeurIPS 2022
conference location
New Orleans, United States
conference dates
2022-11-28 - 2022-12-09
language
English
LU publication?
yes
id
89796d5a-48d6-40ed-94d4-15f9b9749045
date added to LUP
2022-09-19 13:36:30
date last changed
2023-02-13 08:56:20
@inproceedings{89796d5a-48d6-40ed-94d4-15f9b9749045,
  abstract     = {{Information-theoretic Bayesian optimization techniques have become popular for optimizing expensive-to-evaluate black-box functions due to their non-myopic qualities. Entropy Search and Predictive Entropy Search both consider the entropy over the optimum in the input space, while the recent Max-value Entropy Search considers the entropy over the optimal value in the output space. We propose Joint Entropy Search (JES), a novel information-theoretic acquisition function that considers an entirely new quantity, namely the entropy over the joint optimal probability density over both input and output space. To incorporate this information, we consider the reduction in entropy from conditioning on fantasized optimal input/output pairs. The resulting approach primarily relies on standard GP machinery and removes complex approximations typically associated with information-theoretic methods. With minimal computational overhead, JES shows superior decision-making, and yields state-of-the-art performance for information-theoretic approaches across a wide suite of tasks. As a light-weight approach with superior results, JES provides a new go-to acquisition function for Bayesian optimization.}},
  author       = {{Hvarfner, Carl and Hutter, Frank and Nardi, Luigi}},
  booktitle    = {{Advances in Neural Information Processing Systems 35 (NeurIPS 2022)}},
  language     = {{eng}},
  month        = {{09}},
  publisher    = {{Curran Associates, Inc}},
  title        = {{Joint Entropy Search for Maximally-Informed Bayesian Optimization}},
  year         = {{2022}},
}