Cross-Modal Prototype based Multimodal Federated Learning under Severely Missing Modality

TL;DR

This paper introduces MFCPL, a novel method for multimodal federated learning that effectively handles severely missing modalities by leveraging cross-modal prototypes and alignment, improving robustness and performance.

Contribution

The paper proposes MFCPL, a new approach that uses cross-modal prototypes and regularization to address missing modalities in federated learning, enhancing model robustness.

Findings

MFCPL outperforms existing methods on three multimodal datasets.

The approach effectively mitigates data heterogeneity and missing modality issues.

Experimental results show improved global model performance and robustness.

Abstract

Multimodal federated learning (MFL) has emerged as a decentralized machine learning paradigm, allowing multiple clients with different modalities to collaborate on training a global model across diverse data sources without sharing their private data. However, challenges, such as data heterogeneity and severely missing modalities, pose crucial hindrances to the robustness of MFL, significantly impacting the performance of global model. The occurrence of missing modalities in real-world applications, such as autonomous driving, often arises from factors like sensor failures, leading knowledge gaps during the training process. Specifically, the absence of a modality introduces misalignment during the local training phase, stemming from zero-filling in the case of clients with missing modalities. Consequently, achieving robust generalization in global model becomes imperative, especially when dealing with clients that have incomplete data. In this paper, we propose $\textbf{Multimodal Federated Cross Prototype Learning (MFCPL)}$, a novel approach for MFL under severely missing modalities. Our MFCPL leverages the complete prototypes to provide diverse modality knowledge in modality-shared level with the cross-modal regularization and modality-specific level with cross-modal contrastive mechanism. Additionally, our approach introduces the cross-modal alignment to provide regularization for modality-specific features, thereby enhancing the overall performance, particularly in scenarios involving severely missing modalities. Through extensive experiments on three multimodal datasets, we demonstrate the effectiveness of MFCPL in mitigating the challenges of data heterogeneity and severely missing modalities while improving the overall performance and robustness of MFL.