Machine Learning for Chemogenomics on HPC in the ExCAPE Project
An overview of the challenges in ExCAPE to use supercomputing efficiently with three key examples dealing with ef-cient ML work, support for multi-task learning using matrix factorization methods and the challenges originating from the large and very sparse datasets in Ex CAPE are given.
Abstract
—The ExCAPE project is a Horizon 2020 project to advance the state of the art of machine learning (ML) implementations on supercomputing hardware. We have adopted bioactivity predictions for chemogenomics as a challenging use-case to drive development. In this paper, we will give an overview of the challenges in ExCAPE to use supercomputing efficiently. We will touch on three key examples dealing with efficient ML workflow execution, support for multi-task learning using matrix factorization methods and the challenges originating from the large and very sparse datasets in ExCAPE.