AISTATS 2024poster0 citations

Pessimistic Off-Policy Multi-Objective Optimization

Shima Alizadeh, Aniruddha Bhargava, Karthick Gopalswamy, Lalit Jain, Branislav Kveton, Ge Liu

Abstract

Multi-objective optimization is a class of optimization problems with multiple conflicting objectives. We study offline optimization of multi-objective policies from data collected by a previously deployed policy. We propose a pessimistic estimator for policy values that can be easily plugged into existing formulas for hypervolume computation and optimized. The estimator is based on inverse propensity scores (IPS), and improves upon a naive IPS estimator in both theory and experiments. Our analysis is general, and applies beyond our IPS estimators and methods for optimizing them.

BibTeX
@InProceedings{pmlr-v238-alizadeh24a,
  title = 	 {Pessimistic Off-Policy Multi-Objective Optimization},
  author =       {Alizadeh, Shima and Bhargava, Aniruddha and Gopalswamy, Karthick and Jain, Lalit and Kveton, Branislav and Liu, Ge},
  booktitle = 	 {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics},
  pages = 	 {2980--2988},
  year = 	 {2024},
  editor = 	 {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen},
  volume = 	 {238},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {02--04 May},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v238/alizadeh24a/alizadeh24a.pdf},
  url = 	 {https://proceedings.mlr.press/v238/alizadeh24a.html},
  abstract = 	 {Multi-objective optimization is a class of optimization problems with multiple conflicting objectives. We study offline optimization of multi-objective policies from data collected by a previously deployed policy. We propose a pessimistic estimator for policy values that can be easily plugged into existing formulas for hypervolume computation and optimized. The estimator is based on inverse propensity scores (IPS), and improves upon a naive IPS estimator in both theory and experiments. Our analysis is general, and applies beyond our IPS estimators and methods for optimizing them.}
}