Multidimensional data visualization guided by machine learning

Abstract

This project will address two machine learning problems: (I) feature learning and (II) active learning. In both cases, the strategy is to involve the expert in the machine learning loop by means of visual analytics tools. In (I), the expert shall be able to understand and intervene in the deep learning process in order to design more effective image descriptors. In (II), the expert shall be able to select key examples for label supervision and design of an image classifier. The diagnosis of intestinal parasites in microscopy images is the main application, but the validation of the techniques will involve other types of images. (AU)