FuseMoE: Mixture-of-Experts Transformers for Fleximodal Fusion

TL;DR

FuseMoE is a novel mixture-of-experts framework with an innovative gating function that effectively handles multimodal data with missing elements and irregular sampling, improving predictive performance across various tasks.

Contribution

Introducing FuseMoE, a mixture-of-experts model with a new gating function that manages multimodal data with missing modalities and irregular sampling, enhancing convergence and performance.

Findings

Improved convergence rates due to the new gating function.

Effective handling of missing modalities and irregular data.

Validated performance on diverse real-world prediction tasks.

Abstract

As machine learning models in critical fields increasingly grapple with multimodal data, they face the dual challenges of handling a wide array of modalities, often incomplete due to missing elements, and the temporal irregularity and sparsity of collected samples. Successfully leveraging this complex data, while overcoming the scarcity of high-quality training samples, is key to improving these models' predictive performance. We introduce ``FuseMoE'', a mixture-of-experts framework incorporated with an innovative gating function. Designed to integrate a diverse number of modalities, FuseMoE is effective in managing scenarios with missing modalities and irregularly sampled data trajectories. Theoretically, our unique gating function contributes to enhanced convergence rates, leading to better performance in multiple downstream tasks. The practical utility of FuseMoE in the real world is validated by a diverse set of challenging prediction tasks.