E-PUR: an energy-efficient processing unit for recurrent neural networks.
Franyell SilfaGem DotJosé-María ArnauAntonio GonzálezPublished in: PACT (2018)
Keyphrases
- recurrent neural networks
- processing units
- wireless sensor networks
- computing systems
- parallel processing
- parallel computing
- neural network
- feed forward
- sensor networks
- recurrent networks
- cascade correlation
- echo state networks
- reservoir computing
- energy efficiency
- energy consumption
- multiple types
- nonlinear dynamic systems
- neural model
- artificial neural networks
- real time
- massively parallel
- graph cuts
- training set