Kernel Operations on the GPU, with Autodiff, without Memory Overflows.
Benjamin CharlierJean FeydyJoan Alexis GlaunèsFrançois-David CollinGhislain DurifPublished in: J. Mach. Learn. Res. (2021)
Keyphrases
- memory usage
- real time
- kernel function
- storage media
- data transfer
- computing power
- memory requirements
- graphics processors
- feature space
- kernel methods
- parallel processing
- neural network
- write operations
- buffer overflow
- gpu accelerated
- kernel regression
- parallel implementation
- main memory
- support vector
- kernel matrix
- parallel computing
- memory bandwidth