Speakers – INISTA 2023

Dimitrios Katsaros

echicha — Sun, 10 Sep 2023 05:55:07 +0000

Dimitrios Katsaros

Associate professor at the University of Thessaly, Greece

Dimitrios Katsaros is associate professor with the department of Electrical and Computer Engineering at the University of Thessaly, Greece. During the spring semester of 2015 (and in the summer of 2017) he was a visiting fellow (a visiting assistant professor, respectively) in the Department of Electrical Engineering at the Yale university, affiliated also with the Yale Institute for Network Science. During the spring semester of 2019, he was a visiting professor in KIOS Research and Innovation Centre of Excellence at the University of Cyprus. He serves in the editorial board of ACM/Springer Wireless Networks, of Wiley/Hindawi Wireless Communications and Mobile Computing, and of Human-centric Computing and Information Sciences; he has co-guest edited special issues in IEEE Internet Computing, IEEE Network and in other periodicals. His research interests lie in the area of distributed computing and systems.

Abstract: Network Science Elements for Ad Hoc Networking and for Deep Learning

Network science is an inter/multi-disciplinary field engaging concepts, techniques methodologies mainly from mathematics/graph theory, computer science, physics, and social sciences. During the past twenty years and due to the unprecedented capability of humanity to collect, store and process tremendous volumes of data, the field is steadily flourishing. Network science has primarily be used for descriptive purposes, e.g., what is the law that governs the growth of a network, or what is the most central node(s) in a network, or what are the most densely connected groups of nodes of a network, and partially for predictive purposes, e.g., what is the next link that will appear in the network? However, network science concepts can be used in a constructive manner for developing distributed or centralized algorithms for addressing fundamental problems in a diverse set of areas. In this talk, after a presentation of recently and not so recently introduced network science concepts, we will show how such concepts can be used in calculating backbones for information
dissemination into (or control of) wireless ad hoc networks. Then, we will show how other concepts can be used to develop size-reduced neural network models so as to improve training and inference time in deep learning tasks with a minimal sacrifice of accuracy.

Tarek Gasmi

echicha — Wed, 30 Aug 2023 08:37:10 +0000

Tarek Gasmi

Assistant Professor at Sousse University, Tunisia

Tarek Gasmi is an Assistant Professor in Computer Science. Tarek’s academic research and industrial expertise encompass Edge and Cloud AI, MLOps, LLMOps and data analytics. Tarek is recognized as Nvidia Faculty Ambassador and a Microsoft Certified Trainer. Additionally, Tarek contributes as a co-founder of an innovative startup, focusing on the powerful combination of Computer Vision and Large Language Models, particularly in the domain of intelligent video analytics.

Abstract: AI-Powered Video Analytics for Building Smarter Physical Spaces

This tutorial delves into the cutting-edge realm of Edge AI and its profound influence on diverse industries. With the ever-growing presence of cameras and sensors in our daily routines, Edge AI is fundamentally altering how we handle and engage with physical resources. From retail establishments to factory premises, educational campuses, medical facilities, and major road networks, Edge AI is empowering cities and businesses to optimize efficiency, enhance safety, and elevate the overall customer experience. We will examine the advantages of integrating AI directly into edge infrastructure, as well as the obstacles that hinder the widespread adoption of AIoT technologies. Moreover, we will showcase a range of methods for deploying and expediting AI at the edge, assess their performance, and quantify the cost-related factors that influence the development of edge AI strategies.

Schahram Dustdar

echicha — Mon, 22 May 2023 14:04:56 +0000

Schahram Dustdar

Head of the Research Division of Distributed Systems at the TU Wien, Austria

Schahram Dustdar is a Full Professor of Computer Science at the TU Wien, heading the Research Division of Distributed Systems, Austria. He holds several honorary positions: Univers ity of California (USC) Los Angeles; Monash University in Melbourne, Shanghai University, Macquarie University in Sydney, University Pompeu Fabra, Barcelona, Spain. From Dec 2016 until Jan 2017 he was a Visiting Professor at the University of Sevilla, Spain and from January until June 2017 he was a Visiting Professor at UC Berkeley, USA.

From 1999 – 2007 he worked as the co founder and chief scientist of Caramba Labs Software AG in Vienna (acquired by ProjectNetWorld AG), a venture capital co-funded software company focused on software for collaborative processes in teams. He is co-founder of edorer.com (an EdTech company based in the US) and co founder and chief scientist of Sinoaus.net, a Nanjing, China based R&D organization focusing on IoT and Edge Intelligence.

He serves as Editor-in-Chief of Computing (Springer). Dustdar is recipient of multiple awards: IEEE TCSVC Outstanding Leadership Award (2018), IEEE TCSC Award for Excellence in Scalable Computing (2019), ACM Distinguished Scientist (2009), ACM Distinguished Speaker (2021), IBM Faculty Award (2012). He is an elected member of the Academia Europaea: The Academy of Europe, as well as an IEEE Fellow(2016) and an Asia-Pacific Artificial Intelligence Association (AAIA) Fellow (2021) and the AAIA president (since 2021).

Abstract: Distributed Intelligence in the Computing Continuum

Modern distributed systems also deal with uncertain scenarios, where environments, infrastructures, and applications are widely diverse. In the scope of IoT-Edge-Fog-Cloud computing, leveraging these neuroscience-inspired principles and mechanisms could aid in building more flexible solutions able to generalize over different environments.A captivating set of hypotheses from the field of neuroscience suggests that human and animal brain mechanisms result from few powerful principles. If proved to be accurate, these assumptions could open a deep understanding of the way humans and animals manage to cope with the unpredictability of events and imagination.

Eric Gaussier

echicha — Sat, 21 Jan 2023 10:55:24 +0000

Eric Gaussier

Professor at Université Grenoble Alpes, France

After a PhD, conducted jointly at the scientific center of IBM France in Paris and Université Paris 7, and a research and teaching assistant position at Université Paris 7, I joined the European research center of Xerox in Meylan in 1996. I spent six months at the Palo Alto Research Center (PARC, California) when I was employed by Xerox. I finally joined the Université Joseph Fourier, now Université Grenoble Alpes, in Grenoble as a professor in computer science in 2006. I was Director of the Grenoble Computer Science Laboratory (LIG) from 2016-2020, and am, since Sept. 2019, Director and Scientific Director of the Grenoble Interdisplinary Institute in Artificial Intelligence.

Abstract: Overcoming some of the limitations of modern LLMs (large language models)

Tentative abstract: If LLMs behave very well on in-distribution datasets, they fail to generalize to out-of-distribution datasets, so that they cannot be deployed on data which differ from the datasets they have been trained on. Furthermore, they rely on processes which are costly both in terms of data annotation and computation, so that it is difficult to, e.g., deploy them on long documents. After illustrating some important limitations of LLMs, we’ll present and discuss the solutions that have been envisaged so far.

Dimitrios Tzovaras

echicha — Fri, 25 Nov 2022 08:09:43 +0000

Dimitrios Tzovaras

Chairman of the Board of Directors of the Centre for Research and Technology Hellas – CERTH, Greece

Dr. Dimitrios Tzovaras is the Director of Central Directorate and Chairman of the Board of Directors of CERTH since December 2020. Prior to his current position and since October 2013 he was the Director of the Information Technologies Institute of CERTH (CERTH/ITI), while he remains the Head of the Virtual and Augmented Reality Laboratory of CERTH/ITI. He received the Diploma in Electrical Engineering (1992) and the Ph.D. in “2D and 3D Image Compression” (1997) from the Aristotle University of Thessaloniki and holds the position of Researcher A’ of CERTH since 2010.

For eight years (2010-2018) he was a Visiting Professor at Imperial College London (Faculty of Engineering, Department of Electrical and Electronic Engineering, Intelligent Systems & Networks group), while currently he is a Visiting Professor at the University of Nicosia (Institute For the Future – IFF).

His main research interests include network and visual analytics, computer vision, virtual and augmented reality, , multidimensional data analysis, machine learning and artificial intelligence. He is also deeply engaged in applied research aiming at providing innovative solutions in the fields of Energy, Medicine, Industry 4.0, Transport and Culture.

Dr. Tzovaras has conducted very important scientific and research work, which is summarized in 3 books, 55 book chapters, 245 publications in International Journals with Referees and 625 presentations in International Conferences with Referees. Dr. Tzovaras has acted as a reviewer of a large number of submitted scientific papers for a plethora of International Journals and Magazines, and International Scientific Conferences, and has

been Associate Editor of “EURASIP Journal of Applied Signal Processing (JASP)” and “Computer Journal” and Senior Associate Editor of IEEE’s international acknowledged scientific journals “Transactions on Image Processing” and “Signal Processing Letters”.

Since 1992, he has participated with his team in more than 280 Research and Development projects funded by the European Commission and the Greek Secretariat for Research and Innovation and has concluded industrial contracts with leading companies in the world, such as Samsung. He has an extensive management record having been the project coordinator and/or the technical/ scientific manager of 55 projects, during the last 5 years.

Since 2007 he is involved in various activities, management committees and councils relevant to the definition of R&D policy for both the European Commission and the National Agencies (elected member of the National Representatives of Greece in the European Commission for the FP7 Programme of International Cooperation, a member of Board of the National Documentation Centre (EKT) for a three-year term, since March 2014).

Dr Dimitrios Tzovaras is actively involved in innovation and entrepreneurship initiatives. He is the co-founder of nine (9) spin-offs of CERTH/ITI and the initiator of one of the most innovative activities in the area of electronically equipped infrastructures in Greece, the near zero Energy Smart House, an officially recognized Digital Innovation Hub of the EU. He is also the coordinator of Mega Project 3, Artificial Intelligence and Simulation Applications of the 4th Generation Technologic Park, ThessInTec in Northen Greece. He is the owner of seven patents in the scientific areas of virtual reality and health and biometrics, whereas he has applied for two new patents related to Information and Communication Technologies (ICT).

Abstract: AI for trustworthy Breaking News management

Breaking news refers to the information received regarding an event that has just occurred and/or is currently developing. In the era of the Internet of Everything facilitated by the beyond 5G technologies of superfast communication, breaking news are usually escorted by vast amounts of unfiltered, multi-modal information and are digitally broadcasted massively within short time periods. This raises a series of challenges that can be affecting crucial economic sectors, ranging from stock markets to financial products like insurances funds, as well as socio-ethical aspects, from politics to public opinion formulation.

From a more scientific perspective, it can be said that we are dealing with a severe big data challenge, where the unstructured information needs to be carefully collected & efficiently processed, in order to facilitate the utmost scope of trustworthy knowledge extraction. Within this respect, this session discusses a 3-pillar approach that looks prominent to support an AI-enabled pipeline towards the accurate, unbiased & democratic management of Breaking News.

The 1^st pillar implements smart social sensing technologies in order to serve for the detection, identification & collection of the scattered emerging information. Specifically, it utilizes content-aware multi-modal parsers that are able to apply keyword-less NLP-based, topic detection on social media streams, digital news sites & further up-to-date open sources (e.g. Copernicus), that identify the spatio-temporal identity of emerging news, through a dedicated news aggregation engine.

The 2^nd pillar regards the interpretation of the collected & non-documented information. Despite the fact that breaking news are, in their majority, expressed in written form, they often come along (or are correlated) with multimodal attachments (e.g., pictures, satellite images, videos, etc.). One the one hand, multimodal items allow for the transfer of dense information, but on the other hand they further increase the dimensionality of the data to be processed. Thus, in order to lay the focus on the useful information, multimodal AI technologies (e.g. key-frame-extraction, segmentation, detection, annotation, etc.) are applied, capable of extracting only the significant analytics that strengthen the validity and/or the content of the message communicated.

Finally, the 3^rd pillar focuses on the credibility of the identified information, in order to enhance the trustworthiness of the framework. From one perspective, the complicated nature of news propagation can be modelled as a constantly evolving network, the participants (i.e. nodes) of which are simultaneously acting as news generators, (un-)intentional proxies and/or news recipients. This way, we can associate the network as a dynamic graph, where the credibility of the transferred packets (i.e. breaking news) is as good as the profile of the proxies they have passed so far. To this direction, a novel graph-based approach for online evaluation of the validity of emerging news will be explained.

Mirjana Ivanovic

echicha — Thu, 08 Sep 2022 14:54:35 +0000

Mirjana Ivanovic

Full Professor at Faculty of Sciences, University of Novi Sad, Serbia

Mirjana Ivanovic holds the position of Full Professor at Faculty of Sciences, University of Novi Sad, Serbia. She is a member of National Scientific Committee for Electronics, Telecommunication and Informatics within Ministry of Education, Science and Technological Development, Republic of Serbia and member of Board of directors of the Institute for Artificial Intelligence Research and Development of Serbia. She was a member of University Council for Informatics for more than 12 years. Prof. Ivanovic is author or co-author of 14 textbooks, several international monographs and more than 450 research papers, most of which are published in international journals and conferences. Her research interests include agent technologies, intelligent techniques, applications of data mining and machine learning techniques in medical domains and technology enhanced learning. She is member of Program Committees of more than 300 international conferences, Program/General Chair of several international conferences, and leader of numerous international research projects. Mirjana Ivanovic delivered numerous keynote speeches at international conferences and visited many academic institutions all over the world as visiting researcher (Germany, Slovenia, Portugal, Australia, China, Korea). Currently she is Editor-in-Chief of the Computer Science and Information Systems journal.

Abstract: Role of Federated Learning Paradigm in Modern Medical and Health Systems

Federated Learning (FL) is contemporary distributed machine learning paradigm based on a collaboratively decentralized privacy-preserving technology to overcome challenges of data silos and data sensibility. It refers to multiple clients (such as mobile devices, institutions, organizations, etc.) coordinated with one or more central servers for decentralized machine learning settings. FL is appropriate for application when data are privacy-sensitive. Among wide range of possible domains where application of FL can bring additional value are medicine and health care. Each medical institution might have a lot of patient data, but it can be not enough to train their own prediction models. According to that, the combination of FL and prediction of future patients’ status but also patients’ treatment for achieving satisfactory quality of life parameters is good solutions to break down the barriers of analysis throughout different hospitals.

Concerning medical and health domains FL can be employed in different ways like: Use specific models for phenotyping analysis to obtain information concealed in health records without sharing patient-level data. Processing patients’ data exploring different privacy preserving approaches on Electronic Medical Record in federated setting without transmission of data among hospitals’ databases. To detect similar patients scattered in different hospitals without sharing patient-level information and based on obtained predictive models to support medical decisions. Models trained by FL can achieve performance level comparable to ones trained on centrally hosted datasets and can be superior to models that only see isolated single-institutional data.

In this presentation we will focus on crucial aspects of FL paradigm and present its essential characteristics especially connected to medical domains and considering several privacy preserving techniques employed within FL approaches. We will also illustrate effects and power of this recent learning paradigm presenting the vital functionalities of several characteristic medical systems.

Additionally, we will also briefly introduce some promising directions to lead future improvement in use of FL in medical and health domains.