Compressing Transformers: Features Are Low-Rank, but Weights Are Not!
Hao YuJianxin WuPublished in: AAAI (2023)
Keyphrases
- low rank
- linear combination
- convex optimization
- matrix completion
- missing data
- rank minimization
- feature analysis
- singular value decomposition
- feature vectors
- collaborative filtering
- matrix factorization
- feature extraction
- neural network
- higher order
- high order
- kernel matrix
- weight vector
- singular values
- face recognition
- image processing
- low rank matrix
- trace norm