Parallel Implementation of Fuzzy Joint Points Clustering Algorithm on GPU

Fuzzy joint points (FJP) is a fully unsupervised neighborhood-based clustering method that uses a fuzzy neighborhood relationship and overcomes the parameter selection problem of classical neighborhood based clustering algorithms. The present work introduces a parallel implementation of the FJP algorithm on GPU using CUDA in order to reducing the processing time. Provided experimental results confirm a speed up of around 8 times over serial implementation is achieved in the GPU-parallel implementation of the FJP algorithm. So, the work shows that the GPU implementation of the FJP algorithm is a viable option if a speedup is needed.