The research carried out at CiTIUS in the field of machine learning aims to design and develop models, methods and technologies that allow the computer to incorporate new capabilities based on the identification of regularities in observation data. The challenge lies in making learning possible from the availability of very little data to the progressive access to huge volumes of data, assuming that these data can take different forms of representation: image, video, text, etc. It is important to strike a balance between plasticity and generalisability, making efficient use of computational resources, minimising their environmental impact and maximising their economic viability. The application of machine learning is extensive to any socio-economic sector and to any scientific, social or humanistic discipline.
On the other hand, research is being carried out in the search for models, methods and techniques to provide computers with the ability to reason, making it possible to compare previously acquired knowledge with the evidence available in domains that evolve over time, in order to reach conclusions that are useful for decision-making. The challenge lies primarily in formalising the basis of reasoning that underpins rationality, based on information that may be vague, imprecise or uncertain. The application domains are multiple, ranging from health, industry, agri-food, transport or logistics, among others.