Weighted Quantization-Regularization in DNNs for Weight Memory Minimization Toward HW Implementation.
Matthias WessSai Manoj Pudukotai DinakarraoAxel JantschPublished in: IEEE Trans. Comput. Aided Des. Integr. Circuits Syst. (2018)
Keyphrases
- rapid prototyping
- total weight
- efficient implementation
- memory management
- memory requirements
- half quadratic
- regularization term
- objective function
- weight function
- memory space
- random access
- image restoration and reconstruction
- quantization error
- development environment
- hardware implementation
- main memory
- computational complexity