Cellular Neural Networks simulation on a parallel graphics processing unit

The last generation of graphics cards hosts an array of processors, which can be efficiently employed to simulate the cellular neural networks dynamic. In this paper we show an implementation done using CUDA (compute unified device architecture), a language expressively created for architectures with a multiprocessor core. We compare the results with an optimized CNN implementation on CPU, showing how the current graphics cards technology can be competitively applied to this field.