Scalable Pareto Front Approximation for Deep Multi-Objective Learning

TL;DR

This paper introduces a scalable method for approximating the Pareto front in deep multi-objective learning by conditioning networks on preferences, achieving high-quality solutions efficiently.

Contribution

It proposes a novel approach that conditions neural networks directly on preferences and maintains a well-spread Pareto front, significantly reducing training overhead.

Findings

Achieves state-of-the-art Pareto front quality

Computes Pareto fronts significantly faster

Scales to large networks with minimal overhead

Abstract

Multi-objective optimization (MOO) is a prevalent challenge for Deep Learning, however, there exists no scalable MOO solution for truly deep neural networks. Prior work either demand optimizing a new network for every point on the Pareto front, or induce a large overhead to the number of trainable parameters by using hyper-networks conditioned on modifiable preferences. In this paper, we propose to condition the network directly on these preferences by augmenting them to the feature space. Furthermore, we ensure a well-spread Pareto front by penalizing the solutions to maintain a small angle to the preference vector. In a series of experiments, we demonstrate that our Pareto fronts achieve state-of-the-art quality despite being computed significantly faster. Furthermore, we showcase the scalability as our method approximates the full Pareto front on the CelebA dataset with an EfficientNet network at a tiny training time overhead of 7% compared to a simple single-objective optimization. We make our code publicly available at https://github.com/ruchtem/cosmos.