PSB-RNN: A Processing-in-Memory Systolic Array Architecture using Block Circulant Matrices for Recurrent Neural Networks.
Nagadastagiri ChallapalleSahithi RampalliMakesh ChandranGurpreet S. KalsiSreenivas SubramoneyJohn SampsonVijaykrishnan NarayananPublished in: DATE (2020)
Keyphrases
- recurrent neural networks
- systolic array
- parallel architecture
- processing elements
- reconfigurable architecture
- parallel processing
- hardware implementation
- data flow
- memory management
- neural network
- long short term memory
- recurrent networks
- feed forward
- shared memory
- cascade correlation
- artificial neural networks
- neural model
- reservoir computing
- echo state networks
- computational power
- coefficient matrix
- feedforward neural networks
- distributed memory
- least squares
- parallel implementation
- nonlinear dynamic systems
- artificial intelligence
- real time
- chaotic time series
- image segmentation
- efficient implementation
- associative memory