Online Learning of Multiple Tasks and Their Relationships : Testing on Spam Email Data and EEG Signals Recorded in Construction Fields

TL;DR

This paper proposes an online multi-task learning framework that dynamically updates task relatedness, demonstrating improved accuracy on spam email and EEG datasets compared to fixed relatedness methods.

Contribution

The study introduces a novel online multi-task learning approach that iteratively updates task relatedness, unlike previous models assuming static relationships.

Findings

OMTL methods outperform fixed relatedness models by 1-3% accuracy.

Achieved around 12% error rate on spam dataset.

Enhanced performance on EEG datasets under varying conditions.

Abstract

This paper examines an online multi-task learning (OMTL) method, which processes data sequentially to predict labels across related tasks. The framework learns task weights and their relatedness concurrently. Unlike previous models that assumed static task relatedness, our approach treats tasks as initially independent, updating their relatedness iteratively using newly calculated weight vectors. We introduced three rules to update the task relatedness matrix: OMTLCOV, OMTLLOG, and OMTLVON, and compared them against a conventional method (CMTL) that uses a fixed relatedness value. Performance evaluations on three datasets a spam dataset and two EEG datasets from construction workers under varying conditions demonstrated that our OMTL methods outperform CMTL, improving accuracy by 1% to 3% on EEG data, and maintaining low error rates around 12% on the spam dataset.