Collaborative Learning via Prediction Consensus

TL;DR

This paper introduces a collaborative learning method where agents improve their models by sharing pseudo-labeled auxiliary data, using trust weights to reach consensus and enhance performance while reducing communication costs.

Contribution

A novel distillation-based collaborative learning approach with adaptive trust weighting that handles heterogeneity and minimizes negative influence from poor models.

Findings

Significant performance improvements in target domain models.

Effective handling of heterogeneous model architectures.

Reduced communication overhead compared to traditional methods.

Abstract

We consider a collaborative learning setting where the goal of each agent is to improve their own model by leveraging the expertise of collaborators, in addition to their own training data. To facilitate the exchange of expertise among agents, we propose a distillation-based method leveraging shared unlabeled auxiliary data, which is pseudo-labeled by the collective. Central to our method is a trust weighting scheme that serves to adaptively weigh the influence of each collaborator on the pseudo-labels until a consensus on how to label the auxiliary data is reached. We demonstrate empirically that our collaboration scheme is able to significantly boost the performance of individual models in the target domain from which the auxiliary data is sampled. By design, our method adeptly accommodates heterogeneity in model architectures and substantially reduces communication overhead compared to typical collaborative learning methods. At the same time, it can provably mitigate the negative impact of bad models on the collective.