At the Intersection of Data and HPC

In this talk, I will go through two research problems to introduce the interplay between machine learning and HPC from a student's perspective. The first part will be our recent work on rank-2 nonnegative matrix factorization for hierarchical clustering. This work shows the importance of memory access pattern in the development of machine learning algorithms/codes. The second part will be on the acceleration of sparse-dense matrix multiplication on the GPU. This is the work during my summer internship this year (still ongoing), so I will focus on its background and motivation.