MMANet：针对不完整多模态学习的间隔感知蒸馏和模态感知正则化（英文版）

英文标题：MMANet: Margin-aware Distillation and Modality-aware Regularization for Incomplete Multimodal Learning中文摘要：本文提出了一种名为 MMANet 的通用框架，用于不完整多模式学习，其中包含三个组件：用于推理的部署网络、将综合多模式信息传输到部署网络的教师网络以及引导部署网络平衡弱模态组合的正则化网络。英文摘要：Multimodal learning has shown great potentials in numerous scenes andattracts increasing interest recently. However, it often encounters the problemof missing modality data and thus suffers severe performance degradation inpractice. To this end, we propose a general framework called MMA