Navigating Conflicting Views: Harnessing Trust for Learning

TL;DR

This paper introduces a trust-based method to improve multi-view classification by accounting for view reliability, leading to more accurate and reliable predictions in real-world scenarios.

Contribution

It proposes a novel computational trust mechanism that enhances the Evidential Multi-view framework by dynamically weighting views based on their reliability.

Findings

Improved accuracy and agreement metrics on six real-world datasets.

Effective conflict resolution through trust-based view weighting.

Scalability demonstrated on large-scale datasets.

Abstract

Resolving conflicts is critical for improving the reliability of multi-view classification. While prior work focuses on learning consistent and informative representations across views, it often assumes perfect alignment and equal importance of all views, an assumption rarely met in real-world scenarios, as some views may express distinct information. To address this, we develop a computational trust-based discounting method that enhances the Evidential Multi-view framework by accounting for the instance-wise reliability of each view through a probability-sensitive trust mechanism. We evaluate our method on six real-world datasets using Top-1 Accuracy, Fleiss' Kappa, and a new metric, Multi-View Agreement with Ground Truth, to assess prediction reliability. We also assess the effectiveness of uncertainty in indicating prediction correctness via AUROC. Additionally, we test the scalability of our method through end-to-end training on a large-scale dataset. The experimental results show that computational trust can effectively resolve conflicts, paving the way for more reliable multi-view classification models in real-world applications. Codes available at: https://github.com/OverfitFlow/Trust4Conflict