MTLComb: multi-task learning combining regression and classification tasks for joint feature selection

TL;DR

This paper introduces MTLComb, a multi-task learning framework that effectively combines regression and classification tasks with a novel loss weighting scheme, improving joint feature selection accuracy in biomedical applications.

Contribution

The paper proposes a provable loss weighting scheme and an MTL algorithm, MTLComb, for joint feature selection across mixed task types, addressing bias issues in multi-task learning.

Findings

Mitigates bias in feature selection for mixed tasks

Demonstrates effectiveness on simulated and biomedical data

Provides a comprehensive software package for implementation

Abstract

Multi-task learning (MTL) is a learning paradigm that enables the simultaneous training of multiple communicating algorithms. Although MTL has been successfully applied to ether regression or classification tasks alone, incorporating mixed types of tasks into a unified MTL framework remains challenging, primarily due to variations in the magnitudes of losses associated with different tasks. This challenge, particularly evident in MTL applications with joint feature selection, often results in biased selections. To overcome this obstacle, we propose a provable loss weighting scheme that analytically determines the optimal weights for balancing regression and classification tasks. This scheme significantly mitigates the otherwise biased feature selection. Building upon this scheme, we introduce MTLComb, an MTL algorithm and software package encompassing optimization procedures, training protocols, and hyperparameter estimation procedures. MTLComb is designed for learning shared predictors among tasks of mixed types. To showcase the efficacy of MTLComb, we conduct tests on both simulated data and biomedical studies pertaining to sepsis and schizophrenia.