Adaptive Modality Distillation for Separable Multimodal Sentiment Analysis

Multimodal sentiment analysis has increasingly attracted attention since with the arrival of complementary data streams, it great potential to improve and go beyond unimodal analysis. In this article, we present an efficient separable multimodal learning method deal tasks modality missing issue. method, tensor is utilized guide evolution each separated representation. To save computational expense, Tucker decomposition introduced, which leads a general extension low-rank fusion more interactions. The in turn, enhances our distillation processing. Comprehensive experiments on three popular datasets, CMU-MOSI, POM, IEMOCAP, show superior performance especially when only partial modalities are available.