David Sossa ● Profesor Asistente

The research project "Matrices, Optimization, and Randomness with Applications in Data Science (MORA-DataS)" is composed of research teams from Bolivia, Chile, France, and Peru. This project is funded by the regional program MATH-Amsud in cooperation with UMSA (Bolivia), ANID, CMM (Chile), MEAE (France), and CONCYTEC (Peru). The aim of this project is to study diverse optimization models, deterministic and stochastic, and to investigate various problems in matrix analysis with potential applications in data science. Some of the research topics are computing angles between convex cones, inverse eigenvalue problems, proximal algorithms for symmetric cone optimization, nonlinear second-order cone programming problems, nonsmooth joint chance constrained optimization problems, and Euclidean Jordan algebras for optimization.

The project deals with commutation principles in Euclidean Jordan Algebras, Normal Decomposition systems and Fan-Theobald-von Newman systems. It propose to deal with the generalization of these principles and the application to variational analysis and the Marcus-de Oliveira determinantal conjecture.

The field of remote sensing is experiencing an unprecedented acceleration. Besides the large public programs such as Sentinel (see e.g. https://sentinel.esa.int/web/sentinel/missions/sentinel-2), private actors are creating fleets of micro-satellites capable of monitoring of the earth with daily revisits. This abundant and cheap data is creating opportunities for developing novel applications for the monitoring of industrial and agricultural activity. The automatic exploitation of this data is bound to specific application domain knowledge, which requires a mastery of advanced techniques such as computer vision and machine learning, as well as expert knowledge in the field of agriculture. To do this, the team must master earth observation satellites, be able to define the adequate mathematical detection theories, and build on a deep knowledge of satellite image processing, while also including expert knowledge in agriculture. This project aims at uniting competences across the fields of computer vision and machine learning, remote sensing to address emerging applications in agronomy. This project will in addition foster the creation of reproducible research by adopting a reproducible research methodology thus contributing the resulting algorithms to the journal Image Processing On-Line (IPOL). The IPOL journal is an initiative to establish a clear and reproducible state-of-the-art in the domain of image processing and computer vision.