MiniFloat-NN and ExSdotp: An ISA Extension and a Modular Open Hardware Unit for Low-Precision Training on RISC-V Cores.
Luca BertacciniGianna PaulinTim FischerStefan MachLuca BeniniPublished in: ARITH (2022)
Keyphrases
- training process
- neural network
- hardware architecture
- modular neural networks
- low cost
- nearest neighbor
- feedforward neural networks
- instruction set
- knn
- feed forward neural networks
- low power consumption
- back propagation
- processing units
- training algorithm
- real time
- application specific
- neural network training
- computing power
- hardware and software
- training set
- embedded systems
- control unit
- artificial neural networks
- multi layer perceptron
- computer architecture
- parallel computing
- low power
- hardware design
- bp neural network
- feed forward
- learning algorithm
- instruction set architecture