An Energy Efficient Accelerator for Bidirectional Recurrent Neural Networks (BiRNNs) Using Hybrid-Iterative Compression With Error Sensitivity.
Guocai NanZhengkuan WangChenghua WangBi WuZhican WangWeiqiang LiuFabrizio LombardiPublished in: IEEE Trans. Circuits Syst. I Regul. Pap. (2021)
Keyphrases
- recurrent neural networks
- feed forward
- recurrent networks
- neural network
- feedforward neural networks
- neural model
- echo state networks
- energy efficient
- hidden layer
- image compression
- error rate
- reservoir computing
- complex valued
- sensitivity analysis
- cascade correlation
- wireless sensor networks
- data compression
- hebbian learning
- artificial neural networks
- long short term memory
- compression algorithm
- parallel implementation
- iterative methods
- compression ratio
- nonlinear dynamic systems
- sensor networks
- energy efficiency
- compression scheme
- long term
- knowledge base