/backpropagation_cuda

Backpropagation for neural network training using CUDA from scratch

Primary LanguageC++GNU General Public License v3.0GPL-3.0

Parallel Neural Training

This is an application that trains, runs and validates a neural network on GPU, given a dataset.
The training of the network is done using the backpropagation algorithm.
The parallelization is done using a mix of CUDA, Pthreads and OMP.

The program runs on a machine with CUDA 7+ installed.
To execute it, run:
$ make
$ ./parallel_neural_training