High-assurance software laboratory

Data has become the most valuable resource for the automation and optimization of tasks arising both in the private and public spheres. The proposed research area/project aims at strengthening both the synergy and quality of the current research of Taltech in this area, while significantly enhancing the capabilities of Uni to cooperate with Estonian industry and public sphere by joint work, consultations, continuous and regular education. The focus of the project is on using machine learning for data science: ML, in particular deep learning, has shown the most promise in advancing the capabilities of future software systems and empowering the whole business of software development. The concrete goal is to increase the manpower and competence in machine learning, while enhancing and cooperating with the existing areas of data science like data and rule mining, data semantics and knowledge representation, natural language data queries, data integration, statistics and data management.

The present project aims to develop novel and enhance existing methods of explainable artificial intelligence for the analysis of human motor functions. Pilot studies have demonstrated promising results to support the diagnosis of neurodegenerative diseases. In addition, we plan to extend the area of application from medicine to cognitive development and cognitive fatigue analysis. The integration of the explainer component will provide medical professionals with the necessary transparency of the decisions made by AI. Application in the area of cognitive development to support the school education process. Cognitive fatigue is known to cause severe injuries and serious financial losses. In-depth understanding of this phenomenon and ability to recognise mental fatigue targets to make the work environment safer and reduce monetary and non-monetary losses in the process of work.

This project studies new dependently typed systems suitable for the development and mechanization of programming language semantics. Particular emphasis is given to languages supporting concurrency and non-determinism, such as transition systems and process calculi. Popular proof assistants based on dependent type theory, such as Agda and Coq, are inadequate for the formal verification of the denotational semantics of such languages because of their insufficiently expressive type systems. We address this issue by extending modern type-theoretic frameworks, such as homotopy type theory, with a new class of coinductive types coming from final coalgebras of accessible functors. In denotational semantics, these types are necessary for handling the non-deterministic and continuously-interactive behavior of processes. The resulting more expressive type systems will prove themselves capable of encoding the formal semantics of various languages with concurrency and non-determinism.