I am a Machine Learning Research Scientist specialising in Machine Learning and Pattern Analysis for Science and Engineering. My primary interests are in model-based and interpretable Machine Learning, especially through Bayesian approaches, with a focus on image and spatial data. Main areas of application include the Earth and Engineering Sciences, including Biomedical Engineering.
I have an interdisciplinary educational and research background, which includes Applied and Computational Mathematics and Engineering. I hold a Dipl.-Ing. degree in Rural and Surveying Engineering from the National Technical University of Athens, an M.Sc. in Information Processing and Neural Networks from King’s College London, and a D.Phil. in Engineering Science (with specialisation in Machine Learning) from the University of Oxford. I have held research positions in Applied and Computational Mathematics at Princeton University and in Imaging Science at the University of Oxford.
Earth Observation Science and Engineering, dealing with the accurate measurement of spatial phenomena and processes, captures data with an unprecedented detail and by a variety of sensors. The main goal of my research is to develop novel mathematical and statistical methods in order to deal with the increasingly complex datasets generated by these technologies and improve our understanding of spatial phenomena at multiple scales through data-driven modeling. The overarching mission of my research is to translate observations into physical principles and information that will potentially help domain-experts and stakeholders make informed decisions regarding space.
The mathematical and computational frameworks that support representation, inference, and learning in the context of my research include: Bayesian Modeling and Inference, Sparse and Low Rank Recovery Models, Inverse Problems, Neural Networks, Structural Pattern Recognition, and Network Science. Computational tools for the implementation of those algorithms include Functional and Data-Parallel programming and Geographic Information Systems.
The core of my work has dealt with applications of Machine Learning and Pattern Analysis to Image Analysis, with a focus on various forms of spatio-temporal and relational (especially graph-structured) data. My principal area of interest is Remote Sensing and Photogrammetry. While I tend to focus on image and raster data, such as radar, digital photogrammetric sensors, and hyperspectral imagery, other modalities, such as digital surface models, are also being actively researched. In collaboration with experts studying biological vision, I have also done extensive work in Neuroimaging, especially for the modeling and decomposition of spatio-temporal data from functional magnetic resonance imaging scanners in order to produce spatial maps of brain activity as well as on extracting networks of brain activity.
My second major area of research has focused on machine learning modeling of physical phenomena related to Engineering and the Geo-Sciences, especially Geodesy and Geotechnical Engineering, with a particular emphasis on the understanding and modeling of natural hazards. For this investigation, sensor signals such as GNSS and geotechnical data are explored. There is a long-standing collaboration with geotechnical experts.
An important application of this work is to pressing environmental issues, such as clean energy, land use and transportation, and sustainability. Informed policy decisions will require the distillation of a wealth of information, and modern Machine Learning is uniquely positioned to offer this analysis.
Contact Evangelos at: email@example.com