Accelerate Binarized Neural Networks with Processing-in-Memory Enabled by RISC-V Custom Instructions.
Che-Chia LinChao-Lin LeeJenq-Kuen LeeHoward WangMing-Yu HungPublished in: ICPP Workshops (2021)
Keyphrases
- neural network
- application specific
- auto associative
- real time
- domain specific
- associative memory
- back propagation
- pattern recognition
- computational power
- multi layer
- memory requirements
- neural network model
- information processing
- feed forward
- multilayer perceptron
- self organizing maps
- processing elements
- artificial neural networks
- random access
- computing power
- low cost
- memory size
- hardware architecture
- limited memory
- memory space
- memory usage
- edge detection
- rbf network
- training process
- input image
- fuzzy logic
- recurrent neural networks