Palu: Compressing KV-Cache with Low-Rank Projection.
Chi-Chih ChangWei-Cheng LinChien-Yu LinChong-Yan ChenYu-Fang HuPei-Shuo WangNing-Chi HuangLuis CezeKai-Chiang WuPublished in: CoRR (2024)
Keyphrases
- low rank
- matrix factorization
- linear combination
- convex optimization
- missing data
- low rank matrix
- singular value decomposition
- regularized regression
- rank minimization
- matrix decomposition
- high dimensional data
- matrix completion
- kernel matrix
- semi supervised
- high order
- low rank matrices
- singular values
- data matrix
- trace norm
- computer vision
- robust principal component analysis
- data sets
- nonnegative matrix factorization
- feature extraction
- low rank representation
- neural network