Research Areas ǀ Artificial Intelligence for Science

Icons-03

In these areas, we are addressing the fields of semantic technologies, computational scientific discovery, computational creativity and machine learning for science.

We are working on semantic technologies to support the process of data analysis in the spirit of open science. In addition, we are collaborating on the development of ontological resources for the domain of optimization. We are working on development of ontological descriptions of machine learning algorithms, development of a web-based system for querying multi-label classification datasets and experiments. We are extending the OntoDM-core ontology with a module for representing machine learning algorithms and included a more detailed representation of algorithms, including terms such as hyperparameter, optimization problem, complexity function, etc.

We are also developing the optimization algorithm benchmarking ontology (OPTION) to support benchmarking of algorithms in the domain of optimization. Our ontology provides the vocabulary needed for semantic annotation of the core entities involved in the benchmarking process, such as algorithms, problems, and evaluation measures. It also provides means for automated data integration, improved interoperability, powerful querying capabilities and reasoning, thereby enriching the value of the benchmark data. We are demonstrating the utility of OPTION by annotating and querying a corpus of benchmark performance data from the BBOB workshop data, a use case which can be easily extended to cover other benchmarking data collections.

In the field of computational scientific discovery, a key area of using artificial intelligence for science, we are proposing the use of probabilistic grammars to represent domain knowledge in equation discovery. We are also applying the approach of automated process-based modeling of dynamic systems to different topics.

In the related field of computational creativity, we are developing new methods for the creative generation of natural language text based on a general architecture for computational creativity. We are applying these methods to the domain of weather reports. We are proposing a novel method for automated generation of scientific questions and integrating it into our RoboCHAIR system.

We are also extensively using machine learning for science, considering scientific data from different domains, resulting in publications in both computer science and application domain literature.

 

Projects in the field of Artificial Intelligence for Science:

N2-0236

Intelligent inference system for biological discoveries and its application to cancer research, 01.01.2022 - 31.12.2024, Sašo Džeroski

PARC

Partnership for the Assessment of Risks from Chemicals, 01.05.2022 - 30.04.2029, Sašo Džeroski, Panče Panov

P2-0103

Knowledge technologies, 1.1.2022 - 31.12.2027, Sašo Džeroski
TAILOR

TAILOR

Foundations of Trustworthy AI Integrating Learning, Optimisation and Reasoning, 1.9.2020-31.8.2023, Sašo Džeroski, Dragi Kocev

N2-0128

Automating the Synthesis and Analysis of Scientific Models, 1.10.2019-30.9.2022, Sašo Džeroski

J7-9400

Neuropsychological dysfunctions caused by low level exposure to selected environmental pollutants in susceptible population, 1.7.2018-30.6.2021, Sašo Džeroski
J2-9230

J2-9230

Improving Reproducibility of Experiments and Reusability of Research Outputs in Complex Data Analysis, 1.7.2018-30.6.2022, Panče Panov
RI-SI ELIXIR

RI-SI ELIXIR

Development of research Infrastructure for the international competitiveness of Slovenian RRI space, 1.6.2018-31.08.2021, Sašo Džeroski
TRAIN

TRAIN

Big Data and Disease Models: A Cross-border Platform for Validated Biotech Industry Kits, 25.9.2017-24.3.2020, Sašo Džeroski

P2-0103

Knowledge technologies, 1.1.2015-28.2.2021, Nada Lavrač

P2-0103

Knowledge technologies, 1.1.2009-31.12.2014, Nada Lavrač