AIDAPT partner spotlight
AΤΗΕΝΑ is a research and technology organisation which focus in Informatics, Communication and Computational Sciences, tackling global challenges, addressing local needs, and producing novel and deep technological results with a broad impact on other sciences, industry, and society at large. ATHENA operates in 6 Greek cities (Athens, Patras, Xanthi, Chania, Tripoli & Mytilene), and its scope includes informatics, data science, robotics, automation, signal processing, artificial intelligence, networking and digital communication, and modelling. Its technologies cover applications in the fields of engineering, computational linguistics, medicine, biology, biodiversity, earth observation, space science, archaeology etc. By participating in more than 450 national and European projects in the last 5 years and producing more than 1000 publications in the same time frame, ATHENA has gained a strong EU reputation as a center of excellence in IT.
About the team involved
ATHENA participates in AI-DAPT via the Information Management Systems Institute (IMSI). IMSI is today one of Greece’s premier research institutes in the areas of large-scale Information Systems and Big Data management, and conducts cutting-edge scientific research and exploit research results in the development of novel core technologies in the following areas: (a) Big Data Analytics and Machine Learning, (b) Distributed and Web Information Systems, (c) Cloud Platforms and Data Services, (d) User-Centric Systems and Applications, and (e) Big Data Research Infrastructures. Our AI-DAPT team has a strong focus on Big Data and Scalable Data Analytics, Machine Learning and Knowledge Discovery. We investigate ML models to make predictions at large scale, involving the learning of new representations from raw data. Key challenges addressed by our team’s activities include the scale and diversity of data, and the explainability/interpretability of AI systems. Examples include fairness-aware explainability methods, explainability methods specific for recommender systems, explainability of AI experimentation pipelines, and post-hoc explainability methods for black-box ML models. Results are applicable to several sectors including health, environment and materials, and geospatial applications.
ATHENA’s role in AI-DAPT
Our key role is to set up the core foundation layer of hybrid, science-guided AI models (WP3), which is a key concept in AI-DAPT. Science-guided models (or first-principle models) have been widely used in the past to develop process models in industrial plants for several domains (manufacturing, robotics, power plants, biomed etc.). They involve complex systems of mathematical differential/algebraic equations to capture process knowledge and domain expertise (i.e., physical and chemical properties, static and dynamic behaviors, causal relationships among observed quantities, etc.), and to support predictive control, process simulation and operational optimization. Those models are by definition white-box models, since they uncover the process’s inner logic and the decision steps involved. Hybrid science-guided models are a combination of science-guided models and ML models. Such a combination may range from loose-coupling scenarios, where an ML model can be trained to correct the output of the science-guided model itself, to tight-coupling scenarios, where the ML follows a specific architecture to incorporate pre-existing knowledge that can be provided by a domain expert.
Moreover, we are also heavily involved in setting up the AI-DAPT research agenda, based on the AI-DAPT scientific and technical objectives and the use-case requirements (WP1). Finally, we are involved in tasks related to data collection, curation, and augmentation, preparing datasets in a consistent and reliable manner to become suitable for driving the AI-DAPT data pipelines (WP2), and in the AI-DAPT core platform design and implementation (WP4). Finally, ATHENA is the Project Coordinator the AI-DAPT project.
Our ambitions
AI-DAPT aims at reinstating the pure data-related work in its rightful place in AI and at reinforcing the generalizability, reliability, trustworthiness and fairness of Al solutions. To this end, it brings forward a data-centric mentality in AI that is effectively fused with a model-centric, science-based approach, across the complete lifecycle of AI-Ops. The aspect of hybrid science-guided AI, coupled with explainable AI methods, is quite challenging for us, and thus a great opportunity to expand our activities towards this direction, since the project involves a number of industrial-scale challenging demonstrators. In ATHENA, we are pursuing a research agenda with emphasis on delivering tangible assets and reusable components and services for data value chains in several domains. Also, through our Digital Innovation Hub and our Technology Transfer Office, we provide start-up and spin-off incubator services, and a one- stop-shop for researchers to gain support, counselling and financial contribution for commercial exploitation of research results. On the research level, Athena will transfer and adapt AI-DAPT technologies in other data domains, where it has a strong research activity. On the commercial level, ATHENA will give the opportunities to the involved researchers to commercialize a full array of services for a data science infrastructure tailored for the manufacture and process industries.
Our key personnel
Theodore Dalamagas is Research Director and Vice Director of Information Management Systems Institute at ATHENA Research Center, as well as co-founder and Research Director of Symbiolabs (spinoff of ATHENA). He received his Diploma in Electrical Engineering from NTU Athens, Greece, his MSc in Advanced Information Systems from Glasgow University, Scotland, and his PhD from NTU Athens. He has more than 20 years of R&I experience of running and coordinating EU and national R&I IT projects. His research and technology areas of interest include: scientific databanks and e-research infrastructures, data Web and information retrieval, data interoperability and integration, as well as data services and data science applications for several domains (e.g., health and genomics, industrial biotechnology, manufacturing). He has published more than 100 articles in international journals and conference proceedings.
Giorgos Giannopoulos received his Diploma in Electrical and Computer Engineering from the National Technical University of Athens (NTUA), Greece, in 2006 and his PhD in Computer Science from NTUA in 2013. He is currently a Scientific Associate at Athena Research and Innovation Center and at the Operational Unit BEYOND Centre of the National Observatory of Athens. His current research interests focus on the adaptation and extension of Machine/Deep Learning methods in various disciplines, including: fairness and explainability of ML algorithms and pipelines; ML-driven next day prediction of forest fires; pattern recognition on RES timeseries; explanation and prediction of Solar events; ML-driven data integration and annotation; and fact checking. He has been involved in more than 20 National and European R&D projects in various roles: software engineer, researcher, work package leader, proposal contributor and coordinator. He has published more than 55 papers and contributed to the development of open source tools in the above areas.
Anargiros Tzerefos received his Bachelor in Informatics and Telecommunications from the University of Peloponnese in 2019, and his MSc in Data Science program (NSCR Demokritos and the University of Peloponnese). Later in 2019, he began working for Information Management Systems Institute of Athena Research Center as a software developer, through which he has participated in many publications. His work mainly revolves around databases, containerized software and NGS (Next Generation Sequencing) data management and has also tutored the databases class in the Department of Informatics & Telecommunication in the University of Peloponnese during the first semester of 2021.
Vasilis Gkolemis completed his Diploma in Electrical and Computer Engineering from the Aristotle University of Thessaloniki (Greece) in 2017, focusing on Deep Learning with applications in Computer Vision. He later worked as a Research Assistant in EU-funded projects at AUTH. He completed his MSc in Data Science at the University of Edinburgh (UK) in 2020, focusing on Probabilistic Machine Learning. Since then, he has been a developer of the software package ELFI, where he implemented the Likelihood-Free Inference method ROMC as part of his dissertation. He is currently a Research Assistant at IMSI of Athena Research and Innovation Center and a Ph.D. candidate in Explainable AI (XAI) at the Harokopio University of Athens. His scientific interests lie in Machine Learning and Deep Learning, specifically in developing explainability techniques quantifying the uncertainty of the explanations.
Eleni Lavasa holds a B.Sc. diploma in Physics, specialized in Astrophysics, Astronomy & Mechanics, from the National & Kapodistrian University of Athens (NKUA). She received her M.Sc. degree in Space Science, Technologies & Applications in 2020, from the University of Peloponnese & the National Observatory of Athens. Since 2020 she has been a Research Assistant at IMSI of Athena Research Center and a PhD candidate for NKUA. Her scientific interest lies mainly in Solar physics & Space Weather, machine learning & deep learning, and explainable AI. Having a strong background in data analysis, in the last few years she has gained experience in the application of XAI methods in both industrial settings and in Space Weather predictive analysis.
Dimitris Skoutas is a Principal Researcher at the Information Management Systems Institute (IMSI) of Athena Research Center. He received his Diploma in Electrical and Computer Engineering from the National Technical University of Athens (NTUA) in 2003, and his PhD in Computer Science also from NTUA in 2008. During 2009-2011, he worked as a post-doctoral researcher at the L3S Research Center in Hannover, Germany. His research interests focus mainly on data integration, exploration and analysis, with emphasis on geospatial and spatio-textual data, time series, and knowledge graphs. He has more than 90 publications in peer-reviewed international journals, conferences and workshops. He is also an associate editor for the GeoInformatica international journal, and a reviewer in several conferences and journals in these areas. He has participated in many EU and national R&D projects, in several of them as principal investigator or project coordinator.