Multi-task learning on the edge: cost-efficiency and theoretical optimality

TL;DR

This paper introduces a distributed multi-task learning algorithm based on supervised PCA that is both theoretically optimal for Gaussian mixtures and computationally efficient, enabling energy savings without sacrificing performance.

Contribution

It presents a novel distributed MTL algorithm based on SPCA that is theoretically optimal for Gaussian mixtures and scalable for practical use.

Findings

Significant energy savings achieved in experiments.

No performance loss observed with the new algorithm.

Validated on synthetic and real benchmark datasets.

Abstract

This article proposes a distributed multi-task learning (MTL) algorithm based on supervised principal component analysis (SPCA) which is: (i) theoretically optimal for Gaussian mixtures, (ii) computationally cheap and scalable. Supporting experiments on synthetic and real benchmark data demonstrate that significant energy gains can be obtained with no performance loss.