https://hgpu.org/?p=7124
Scalability of Self-organizing Maps on a GPU cluster using OpenCL and CUDA