I2MoE: Interpretable Multimodal Interaction-aware Mixture-of-Experts

TL;DR

I2MoE introduces an interpretable multimodal fusion framework that explicitly models diverse interactions between modalities, improving performance and interpretability in multimodal learning tasks.

Contribution

The paper presents I2MoE, a novel end-to-end mixture-of-experts model that explicitly captures and interprets multimodal interactions with weakly supervised learning.

Findings

Improves task performance across medical and general datasets.

Provides local and global interpretability of multimodal interactions.

Flexible integration with various fusion techniques.

Abstract

Modality fusion is a cornerstone of multimodal learning, enabling information integration from diverse data sources. However, vanilla fusion methods are limited by (1) inability to account for heterogeneous interactions between modalities and (2) lack of interpretability in uncovering the multimodal interactions inherent in the data. To this end, we propose I2MoE (Interpretable Multimodal Interaction-aware Mixture of Experts), an end-to-end MoE framework designed to enhance modality fusion by explicitly modeling diverse multimodal interactions, as well as providing interpretation on a local and global level. First, I2MoE utilizes different interaction experts with weakly supervised interaction losses to learn multimodal interactions in a data-driven way. Second, I2MoE deploys a reweighting model that assigns importance scores for the output of each interaction expert, which offers sample-level and dataset-level interpretation. Extensive evaluation of medical and general multimodal datasets shows that I2MoE is flexible enough to be combined with different fusion techniques, consistently improves task performance, and provides interpretation across various real-world scenarios. Code is available at https://github.com/Raina-Xin/I2MoE.