Learning Reliable and Scalable Representations Using Multimodal Multitask Deep Learning

This work enables models to effectively learn fused representations from multiple modalities and across tasks, exploiting complementary features and cross-modal interdependencies and facilitates self-supervised model adaptation.