Deep Belief Nets in C++ and CUDA C: Volume 3: Convolutional Nets

© Timothy Masters 2018

Timothy MastersDeep Belief Nets in C++ and CUDA C: Volume 3https://doi.org/10.1007/978-1-4842-3721-2_3

3. CUDA Code

Timothy Masters¹ 

(1)

Ithaca, New York, USA

 

The source code for the CUDA implementation of convolutional nets is in two files, both of which can be downloaded for free from my web site. MOD_CUDA.CPP provides the high-level organization. It calls subroutines to initialize, compute forward activation, backpropagate delta, and compute the gradient. MOD_CUDA.cu contains the CUDA device routines, as well as the low-level C++ host routines that are called from MOD_CUDA.CPP and that in turn launch computation kernels and provide communication between the host and the device.

Many excellent books on CUDA programming exist. ...