DeMM: A Decoupled Matrix Multiplication Engine Supporting Relaxed Structured Sparsity.
Christodoulos PeltekisVasileios TitopoulosChrysostomos NicopoulosGiorgos DimitrakopoulosPublished in: IEEE Comput. Archit. Lett. (2024)
Keyphrases
- structured sparsity
- matrix multiplication
- message passing
- statistical learning
- compressive sensing
- learning problems
- group lasso
- matrix factorization
- distributed memory
- convex optimization problems
- model selection
- optimal solution
- machine learning
- optimization problems
- high resolution
- image retrieval
- lower bound
- computer vision