Combining Task Predictors via Enhancing Joint Predictability

TL;DR

This paper introduces a Bayesian-based predictor combination algorithm that jointly assesses multiple reference predictors to enhance a target predictor, outperforming existing pairwise methods across diverse real-world datasets.

Contribution

The paper presents a novel Bayesian framework for predictor combination that considers dependencies among references and automatically selects relevant predictors, improving performance.

Findings

Significant performance gains on seven real-world datasets.

Broadens the applicability of predictor combination methods.

Automatically identifies relevant predictors for improved accuracy.

Abstract

Predictor combination aims to improve a (target) predictor of a learning task based on the (reference) predictors of potentially relevant tasks, without having access to the internals of individual predictors. We present a new predictor combination algorithm that improves the target by i) measuring the relevance of references based on their capabilities in predicting the target, and ii) strengthening such estimated relevance. Unlike existing predictor combination approaches that only exploit pairwise relationships between the target and each reference, and thereby ignore potentially useful dependence among references, our algorithm jointly assesses the relevance of all references by adopting a Bayesian framework. This also offers a rigorous way to automatically select only relevant references. Based on experiments on seven real-world datasets from visual attribute ranking and multi-class classification scenarios, we demonstrate that our algorithm offers a significant performance gain and broadens the application range of existing predictor combination approaches.