Implementation of parallel K-means algorithm for image classification using OpenMP and MPI libraries

The k-means algorithm is a widely used Machine learning algorithm for clustering. This paper introduces a parallel k-means algorithm implementation for image classification. We implemented several parallel programs using OpenMP and MPI libraries and tested them on three MNIST variants datasets: MNIST, Fashion MNIST, and Extended MNIST. The algorithm parameters and scheduling methods are varied during experiments. The implementations use both libraries, run on multiple computers, and utilize all available processor cores. The proposed implementations achieve a speedup of up to 10.2 while using 12 processor cores.