Hardware-Aware DNN Compression via Diverse Pruning and Mixed-Precision Quantization.
Konstantinos BalaskasAndreas KaratzasChristos SadKostas SioziosIraklis AnagnostopoulosGeorgios ZervakisJörg HenkelPublished in: CoRR (2023)
Keyphrases
- quantization noise
- lossy image compression
- efficient compression
- data compression
- low cost
- hardware and software
- compression scheme
- uniform quantization
- real time
- image compression
- quantization error
- compression algorithm
- precision and recall
- search space
- pruning method
- huffman coding
- high precision
- bits per pixel
- entropy coding
- transform coding
- real world
- adaptive quantization
- quantization scheme
- lookup table
- bit rate
- wide variety
- computational complexity
- pruning methods
- transform coefficients
- massively parallel
- compression ratio
- training data
- quantization step
- image processing