A distributional view on multi-objective policy optimization
Abbas Abdolmaleki, Sandy Huang, Leonard Hasenclever, Michael Neunert, Francis Song, Martina Zambelli, Murilo Martins, Nicolas Heess
Abstract
Many real-world problems require trading off multiple competing objectives. However, these objectives are often in different units and/or scales, which can make it challenging for practitioners to express numerical preferences over objectives in their native units. In this paper we propose a novel algorithm for multi-objective reinforcement learning that enables setting desired preferences for objectives in a scale-invariant way. We propose to learn an action distribution for each objective, and we use supervised learning to fit a parametric policy to a combination of these distributions. We demonstrate the effectiveness of our approach on challenging high-dimensional real and simulated robotics tasks, and show that setting different preferences in our framework allows us to trace out the space of nondominated solutions.
BibTeX
@InProceedings{pmlr-v119-abdolmaleki20a,
title = {A distributional view on multi-objective policy optimization},
author = {Abdolmaleki, Abbas and Huang, Sandy and Hasenclever, Leonard and Neunert, Michael and Song, Francis and Zambelli, Martina and Martins, Murilo and Heess, Nicolas and Hadsell, Raia and Riedmiller, Martin},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
pages = {11--22},
year = {2020},
editor = {III, Hal Daumé and Singh, Aarti},
volume = {119},
series = {Proceedings of Machine Learning Research},
month = {13--18 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v119/abdolmaleki20a/abdolmaleki20a.pdf},
url = {https://proceedings.mlr.press/v119/abdolmaleki20a.html},
abstract = {Many real-world problems require trading off multiple competing objectives. However, these objectives are often in different units and/or scales, which can make it challenging for practitioners to express numerical preferences over objectives in their native units. In this paper we propose a novel algorithm for multi-objective reinforcement learning that enables setting desired preferences for objectives in a scale-invariant way. We propose to learn an action distribution for each objective, and we use supervised learning to fit a parametric policy to a combination of these distributions. We demonstrate the effectiveness of our approach on challenging high-dimensional real and simulated robotics tasks, and show that setting different preferences in our framework allows us to trace out the space of nondominated solutions.}
}