Gradient-Guided Modality Decoupling for Missing-Modality Robustness
This work solves the challenge of missing-modality robustness in multimodal learning, which is practical for applications like medical segmentation and sentiment analysis, but it is incremental as it builds on existing methods like Grad-CAM.
The paper tackles the problem of multimodal learning with incomplete input data by addressing modality dominance, which degrades performance when modalities are missing, and introduces a Gradient-guided Modality Decoupling method that improves performance on benchmarks like BraTS 2018, CMU-MOSI, and CMU-MOSEI.
Multimodal learning with incomplete input data (missing modality) is practical and challenging. In this work, we conduct an in-depth analysis of this challenge and find that modality dominance has a significant negative impact on the model training, greatly degrading the missing modality performance. Motivated by Grad-CAM, we introduce a novel indicator, gradients, to monitor and reduce modality dominance which widely exists in the missing-modality scenario. In aid of this indicator, we present a novel Gradient-guided Modality Decoupling (GMD) method to decouple the dependency on dominating modalities. Specifically, GMD removes the conflicted gradient components from different modalities to achieve this decoupling, significantly improving the performance. In addition, to flexibly handle modal-incomplete data, we design a parameter-efficient Dynamic Sharing (DS) framework which can adaptively switch on/off the network parameters based on whether one modality is available. We conduct extensive experiments on three popular multimodal benchmarks, including BraTS 2018 for medical segmentation, CMU-MOSI, and CMU-MOSEI for sentiment analysis. The results show that our method can significantly outperform the competitors, showing the effectiveness of the proposed solutions. Our code is released here: https://github.com/HaoWang420/Gradient-guided-Modality-Decoupling.