A 1-TFLOPS/W, 28-nm Deep Neural Network Accelerator Featuring Online Compression and Decompression and BF16 Digital In-Memory-Computing Hardware.
Bo ZhangSeunghyun MoonMingoo SeokPublished in: CICC (2024)
Keyphrases
- neural network
- compression ratio
- data compression
- image compression
- compression scheme
- compressed data
- compression algorithm
- computing power
- low cost
- artificial neural networks
- real time
- circuit design
- internal memory
- random access
- field programmable gate array
- auto associative
- hardware and software
- parallel hardware
- huffman coding
- associative memory
- memory management
- online learning
- back propagation
- memory requirements
- hardware implementation
- main memory
- lossless compression
- camera phones
- computational power
- compute intensive
- neural network model
- massively parallel
- image quality
- data structure
- embedded systems