Where do machine learning and optimization meet?

Abstract

Machine learning (ML) and optimization have provided efficient solutions to many practical problems. Despite several relevant initiatives to bring together researchers from these areas, they still follow independent paths. Many current research challenges would largely benefit from a stronger cooperation among researchers from these areas. This project tries to walk in this direction, by bringing together researchers from machine learning and optimization to explore new venues where knowledge from these two research areas can be combined, resulting in new efficient solutions to relevant applications. From one point of view, optimization techniques can save human and computational resources by exploring different regions of the search space a grid-search (i.e., a commonly used technique to optimize parameters in machine learning) may fail, for instance. On the other hand, machine learning can be a fruitful pool of applications that new optimization techniques can be evaluated and used as a test bed. In this work, we aim at focusing on bioinformatic-related applications, with special attention to breast cancer gene expression analysis, and Autism fMRI data. (AU)