基于MPI的ML-kNN算法并行

Parallelization of ML-kNN Based on MPI

基于MPI将ML-kNN算法并行化,以解决多标签学习领域中的大规模分类问题,控制计算的时间开销,这也是首次将MPI应用到多标签学习领域.通过与传统的串行ML-kNN的对比实验,验证了所提方法的可行性和有效性.另外,允许数据集以特征为单位划分,这使得该方法在处理高维数据时具有更大的优势.

The ML-k NN algorithm was parallelized by MPI(message passing interface).It aimed to illustrate the classification problem of large-scale data and control the computation of overhead time.It was also the first time to apply MPI into multi-label learning area.According to traditional contrast experiments about serial ML-k NN,it verified the feasibility and effectiveness of the algorithm which was provided.It was worthy mentioned that there were some methods which could make the partition of data set due to feature-based unit.It might help people to get more tremendous advantage when dealing with high dimensional data.