search

TmacMai / Multimodal-Information-Bottleneck

Multimodal Information Bottleneck: Learning Minimal Sufficient Unimodal and Multimodal Representations (MIB for multimodal sentiment analysis)
MIT License
54 stars 4 forks source link

MI objective #7

Open LindgeW opened 1 year ago

LindgeW commented 1 year ago

Hi,

Where is the MI loss implemented in your codes? I just found the L1 loss is used instead?