Tensorox: Accelerating GPU Applications via Neural Approximation on Unused Tensor Cores.
Nhut-Minh HoWeng-Fai WongPublished in: IEEE Trans. Parallel Distributed Syst. (2022)
Keyphrases
- tensor product
- neural network
- real time
- approximation error
- approximation algorithms
- higher order
- network architecture
- parallel implementation
- diffusion tensor
- approximation methods
- error bounds
- learning rules
- multi core processors
- parallel architectures
- graphics hardware
- parallel programming
- closed form
- high order
- dimensionality reduction
- parallel computing
- biologically plausible
- pairwise
- diffusion tensor images
- gpu accelerated