Multitask Learning via Shared Features: Algorithms and Hardness
Abstract
We investigate the computational efficiency of multitask learning of Boolean functions over the -dimensional hypercube, that are related by means of a feature representation of size shared across all tasks. We present a polynomial time multitask learning algorithm for the concept class of halfspaces with margin , which is based on a simultaneous boosting technique and requires only samples-per-task and samples in total. In addition, we prove a computational separation, showing that assuming there exists a concept class that cannot be learned in the attribute-efficient model, we can construct another concept class such that can be learned in the attribute-efficient model, but cannot be multitask learned efficiently -- multitask learning this concept class either requires super-polynomial time complexity or a much larger total number of samples.
Cite
@article{arxiv.2209.03112,
title = {Multitask Learning via Shared Features: Algorithms and Hardness},
author = {Konstantina Bairaktari and Guy Blanc and Li-Yang Tan and Jonathan Ullman and Lydia Zakynthinou},
journal= {arXiv preprint arXiv:2209.03112},
year = {2022}
}