Sparse MoE as the New Dropout: Scaling Dense and Self-Slimmable Transformers.
Tianlong ChenZhenyu ZhangAjay JaiswalShiwei LiuZhangyang WangPublished in: CoRR (2023)
Keyphrases
- dense optical flow
- motion field estimation
- dense stereo
- dense reconstruction
- dense motion estimation
- sparse data
- shape recovery
- high dimensional
- compressive sensing
- elastic net
- real time
- optical flow
- multiscale
- compressed sensing
- sparse coding
- dense sampling
- dense disparity map
- sparse representation
- image classification
- object recognition