maxDNN: An Efficient Convolution Kernel for Deep Learning with Maxwell GPUs
eBay Research Labs Machine Learning
arXiv:1501.06633 [cs.NE], (27 Jan 2015)
@article{lavin2015maxdnn,
title={maxDNN: An Efficient Convolution Kernel for Deep Learning with Maxwell GPUs},
author={Lavin, Andrew},
year={2015},
month={jan},
archivePrefix={"arXiv"},
primaryClass={cs.NE}
}
This paper describes maxDNN, a computationally efficient convolution kernel for deep learning with the NVIDIA Maxwell GPU. maxDNN reaches 96.3% computational efficiency on typical deep learning network architectures using a single kernel. The design combines ideas from cuda-convnet2 with the Maxas SGEMM assembly code. We only address forward propagation (FPROP) operation of the network, but we believe that the same techniques used here will be effective for backward propagation (BPROP) as well.
January 28, 2015 by hgpu
