Multi-task problems are not multi-objective
This work challenges the use of MTL as a benchmark for MOO algorithms in machine learning, calling for new evaluation standards, but it is incremental as it critiques existing practices without introducing a new method.
The paper demonstrates that multi-task learning (MTL) problems do not exhibit the competing objectives characteristic of multi-objective optimization (MOO), as a single expressive model can optimize all tasks without trade-offs, making MOO methods inapplicable. This is supported by experiments on Multi-Fashion-MNIST datasets, showing no performance loss compared to independent models.
Multi-objective optimization (MOO) aims at finding a set of optimal configurations for a given set of objectives. A recent line of work applies MOO methods to the typical Machine Learning (ML) setting, which becomes multi-objective if a model should optimize more than one objective, for instance in fair machine learning. These works also use Multi-Task Learning (MTL) problems to benchmark MOO algorithms treating each task as independent objective. In this work we show that MTL problems do not resemble the characteristics of MOO problems. In particular, MTL losses are not competing in case of a sufficiently expressive single model. As a consequence, a single model can perform just as well as optimizing all objectives with independent models, rendering MOO inapplicable. We provide evidence with extensive experiments on the widely used Multi-Fashion-MNIST datasets. Our results call for new benchmarks to evaluate MOO algorithms for ML. Our code is available at: https://github.com/ruchtem/moo-mtl.