Artificial intelligence and inteligent systems

The proposed program will contribute to several aspects of basic research in artificial intelligence (AI), and importantly, also to applications of this technology. Therefore the proposed research program addresses some key questions of AI, in combination with applicative research in corresponding areas. From the scientific perspective, the question of how to achieve or explain various aspects of human intelligence remains open. From application point of view, artificial intelligence has an increasingly important role in the development of information society and rapidly developing sciences like biomedicine. European program Horizon 2020 for ICT includes numerous research directions in science and technology that include AI methods as key ingredients, for example intelligent robotics, factories of the future, e-health, learning and tutoring, energy, transportation, smart platforms/systems/cities, analysis of large collections of data. The proposed research contents is to a large extent a continuation of the program “Artificial intelligence and intelligent systems” in 2009-2014, taking into account some significant recent trends. The research contents is roughly as follows: 1. Machine learning and knowledge discovery in data: Integration of heterogeneous data from various sources, particularly for knowledge discovery in very large data sets; estimating reliability of predictions; approaches with matrix factorisation and archetype analysis; extension of our approach ABML (Argument-Based Machine Learning) to techniques other than rule learning, and optimised selection of critical examples. 2. Qualitative modelling and intelligent robotics: Robot learning with representations of induced models as combination of compositional hierarchies, qualitative models, and numerical constraints; discovery of abstract concepts that are not explicitly present in observation data. 3. Evolutionary computation and heuristic search: evolutionary algorithms for multi criteria optimisation with improved efficiency due to surrogate models and visualisation of results; analysis of pessimistic heuristics in real time search. 4. Intelligent systems for tutoring and commenting: automated teaching with argumentation using the ABML method; assessing problem difficulty for humans; methods for automatic commenting; automated recommendation of teaching materials. 5. Multi-agent systems and ambient intelligence: techniques for integration of various data and knowledge sources, and behaviour analyses adapted to the user, sensors and context; development of a platform for intelligent buildings and cities. 6. Application oriented research: biomedicine (genetics, cancer and neurological disorders), intelligent natural language interfaces, recommender systems, oceanography, etc. 7. Development of tools and program systems: further development of the system Orange for knowledge discovery from data, and domain-specific tools for data analytics; optimisation of ILP system HYPER.

The proposed program envisages research in some of key and timely topics in Artificial Intelligence that are very important due to the impact and applicability of this fast developing area. Therefore we expect that the program will importantly contribute to the development of the selected topics in AI, to practically significant software environments, and confirm the applicability of new scientific results in practice. The questions “What is intelligence?”, and “How it comes about?”, belong to key and insufficiently answered scientific and philosophical questions. This program is contributing also to these questions. In the past, it has already effected philosophical views on the development of science. Currently, our research plan under item 2 (qualitative modelling and intelligent robotics) aims at contributing to one of the fundamental questions of cognitive science: How human knowledge emerges? (J. B. Tenenbaum, et al. How to grow a mind: statistics, structure, and abstraction, Science 331 (6022) (2011) 1279–1285). Here we mean our combined knowledge representation that we are investigating, machine learning using this representation, and mechanisms of automatically discovering new, abstract theoretical concepts. Our proposed research will not be important only for the development of Computer Science. The methods of AI, such as those in machine learning, knowledge discovery from data, the codification and use of domain knowledge, qualitative reasoning and design of intelligent systems, became key ingredients of the infrastructure of other scientific disciplines, in natural and social sciences. The working methods in these sciences have changed due to the availability of AI tools and methods, and the availability of scientific databases. It is now possible to do research in biology, medicine and genetics without necessary availability of a laboratory for experimentation. It is possible to achieve important results in these disciplines just with analysis of publicly available data on the web. The scientists in these disciplines contribute their expertise and skill in the use of data mining methods and tools, possibly in collaboration with AI scientists.  In this, the problem of integration of heterogenous data from various sources is particularly timely and critical.  The proposed research program is largely aimed at this. We intend to continue the development of freely available software that implements methods of data analytics that include also original technical contributions of this research. Our machine learning and data mining system Orange is particularly important in this respect. In addition to numerous techniques of data mining, Orange also includes specialised modules for application in genetics. Orange has very many users world wide, among them are also top scientists from other scientific disciplines. We expect that the research team of this program will, due to excellent connections with user communities and external partners in biology, medicine, genetics and ecological modelling, importantly contribute to AI-based research methods and systems in these areas, and thus also to the development of other scientific disciplines.

Development of information technology and its multidisciplinary use in science, commerce and education is one of the highest national priorities and is listed as such in the corresponding strategic documents. The expected contributions of the proposed program are as follows. Ambiental intelligence, one of the proposed research topics, explores the technical means to enhance the living quality of the aging population of Slovenia and Europe. Connected to this we will continue our work on decision support systems in medicine, especially in the fields of neurology and oncology. Research of language and speech technologies alleviate the use of Slovenian language in high technology products and services, like in more spoken languages, which contributes to preservation of national identity. We will continue to contribute to the advancement of the information society and e-government in Slovenia, among other activities by organizing the Information Society conference, which already has a fifteen year tradition. Methods of evolutionary computation effectively solve search and optimization problems. The proposed research will result in the algorithms capable of improving product quality and energy efficiency of production processes at large industrial manufacturers, which will at the same time enhance their competitiveness and environmental sustainability. Development of automated speaker identification in forensic science will allow the use of modern scientific and technological means for crime investigation, thus increasing the efficiency of the corresponding justice institutions and cutting the court backlogs. A substantial effort will be devoted to development of computational techniques and their implementations for bioinformatics analysis. Our work in crafting data fusion approaches is specifically important for development of data-rich fields like molecular biology, and can have a profound influence on development of this technologically challenging and scientifically and practically extremely important discipline in Slovenia. Many members of the group are university teachers. With the importance of the AI methods and their great potential socio-economic influence, it is important to teach these methods and their practical use to new generations of students. Worth mentioning are two young researchers from previous research programme who received awards for the best European doctoral dissertations in AI in 2002 and 2005. Socio-economic impact will be additionally enriched by our strong collaborations with foreign partners. The group established relations with numerous prestigious foreign institutions and brought a number of foreign scientists on shorter and longer visits to Slovenia. We also plan to involve them in teaching at the University of Ljubljana.