https://hgpu.org/?p=12175
A CUDA-enabled Parallel Implementation of Collaborative Filtering