Towards True Explainable Artificial Intelligence for Real-World Applications
Hani Hagras, University of Essex, United Kingdom
Deep Learning for Active Robotic Perception
Anastasios Tefas, Aristotle University of Thessaloniki, Greece
Next Generation of Multi-Objective Evolutionary Optimization and Decision-Making Algorithms
Sanaz Mostaghim, Otto-von-Guericke-Universität Magdeburg, Germany
Assessment and Evaluation of Empirical and Scientific Data
Nikolaus Hansen, Inria & Ecole Polytechnique & Institut Polytechnique de Paris, France
Towards True Explainable Artificial Intelligence for Real-World Applications
Hani Hagras
University of Essex
United Kingdom
http://cswww.essex.ac.uk/staff/hagras.htm
![]()
Brief Bio
Hani Hagras is a Professor of Artificial Intelligence, Director of Impact, Director of the Computational Intelligence Centre and Head of the Artificial Intelligence Research Group, in the School of Computer Science and Electronic Engineering, University of Essex, UK. He is a Fellow of Institute of Electrical and Electronics Engineers (IEEE), a Fellow of the Institution of Engineering and Technology (IET), Principal Fellow of the UK Higher Education Academy (PFHEA) and Fellow of the Asia-Pacific Artificial Intelligence Association (AAIA)
His main research interests are in Explainable Artificial Intelligence (XAI) and Data Science with applications to Finance, Cyber Physical Systems, Neuroscience, Life Sciences, Uncertainty Management, Intelligent Robotics and Intelligent Control of Industrial Processes.
He has authored more than 500 papers in international journals, conferences and books. He is amongst the top 2% of the most-cited scientists in the World (Scopus August 2021). His work received funding from major research councils and industry. He holds eleven industrial patents in the field of Explainable AI.
His research has won numerous prestigious international awards where he was awarded by the IEEE Computational Intelligence Society (CIS), the 2010 Outstanding Paper Award in the IEEE Transactions on Fuzzy Systems and the 2004 Outstanding Paper Award in the IEEE Transactions on Fuzzy Systems. He was also awarded the 2015 and 2017 Global Telecommunications Business award for his joint project with British Telecom. In 2016, he was elected as Distinguished Lecturer by the IEEE Computational Intelligence Society. His work has also won best paper awards in several leading international conferences including the 2014 and 2006 IEEE International Conference on Fuzzy Systems, the 2012 UK Workshop on Computational Intelligence and the 2016 International Conference of the BCS SGAI International Conference on Artificial Intelligence. He was awarded by the IEEE Computational Intelligence Society (CIS) the 2011 IEEE CIS Outstanding Chapter Award. In 2017, he was awarded by the University of Essex, the 2017 best Research impact award for his work with British Telecom. He acted as the Principal Investigator for a project which was awarded by the UK Technology Strategy Board, the 2011 UK Best Knowledge Transfer Partnership for London and the East Region. He also acted as the Principal Investigator for a project which was awarded the 2009 Lord Stafford Achievement in Innovation Award for East of England. In 2010, he Led a Research Students team to win the First place in the RoboCup 2010. In 2007, he was Shortlisted by the Times Higher Education supplement (THES) for the UK Young researcher of the year award.
He is Associate Editor of many journals including IEEE Transactions on Fuzzy Systems, IEEE Transactions on Artificial Intelligence, Knowledge Based Systems, Cognitive Computations and others.
He served as the General and Programme Chair of numerous major international conferences where he served as the General co-Chair of the 2007 IEEE International Conference on Fuzzy Systems, and Programme Chair of the 2021 and 2017 IEEE International Conference on Fuzzy Systems as well as many other conferences
Abstract
The recent advances in computing power coupled with the rapid increases in the quantity of available data has led to a resurgence in the theory and applications of Artificial Intelligence (AI). However, the use of complex AI algorithms could result in a lack of transparency to users which is termed as black/opaque box models. Thus, for AI to be trusted and widely used by governments and industries, there is a need for greater transparency through the creation of human friendly explainable AI (XAI) systems. XAI aims to make machines understand the context and environment in which they operate, and over time build underlying explanatory models that allow them to characterize real-world phenomena. The XAI concept provides an explanation of individual decisions, enables understanding of overall strengths and weaknesses, and conveys an understanding of how the system will behave in the future and how to correct the system’s mistakes. In this keynote speech, Hani Hagras introduce the concepts of XAI by moving towards “explainable AI” (XAI) to achieve a significantly positive impact on communities and industries all over the world and will present novel techniques enabling to deliver human friendly XAI systems which could be easily understood, analysed and augmented by humans. This will allow to the wider deployment of AI systems which are trusted in various real world applications.
Deep Learning for Active Robotic Perception
Anastasios Tefas
Aristotle University of Thessaloniki
Greece
https://cidl.csd.auth.gr/
![]()
Brief Bio
Anastasios Tefas received the B.Sc. in informatics in 1997 and the Ph.D. degree in informatics in 2002, both from the Aristotle University of Thessaloniki, Greece. Since 2022 he has been a Professor at the Department of Informatics, Aristotle University of Thessaloniki. From 2008 to 2022, he was a Lecturer, Assistant Professor, Associate Professor at the same University. Prof. Tefas participated in 20 research projects financed by national and European funds. He is the Coordinator of the H2020 project OpenDR, “Open Deep Learning Toolkit for Robotics”. He is Area Editor in Signal Processing: Image Communications journal. He has co-authored 150 journal papers, 280 papers in international conferences and contributed 17 chapters to edited books in his area of expertise. He has co-organized more than 15 workshops, tutorials, special sessions and special issues and has given more than 20 invited talks. He has co-edited the book “Deep Learning for Robot Perception and Cognition”, Elsevier, 2022. Over 9600 citations have been recorded to his publications and his H-index is 48 according to Google scholar. His current research interests include computational intelligence, deep learning, pattern recognition, machine learning, digital signal and image analysis and retrieval, computer vision and robotics.
Abstract
Deep Learning (DL) has brought about significant advancements in recent years, greatly enhancing various challenging computer vision tasks. These tasks include but are not limited to object detection and recognition, scene segmentation, and face recognition, among others. DL's advanced perception capabilities have also paved the way for powerful tools in the realm of robotics, resulting in remarkable applications such as autonomous vehicles, drones, and robots capable of seamless interaction with humans, such as collaborative manufacturing. However, despite these remarkable achievements in DL within these domains, a significant limitation persists: most existing methods adhere to a static inference paradigm inherited from traditional computer vision pipelines. Indeed, DL models typically perform inference on a fixed and static input, ignoring the fact that robots possess the capability to interact with their environment to gain a better understanding of their surroundings. This process, known as "active perception", closely mirrors how humans and various animals interact and comprehend their environment. For instance, humans tend to examine objects from different angles, when being uncertain, while some animals have specialized muscles that allow them to orient their ears towards the source of an auditory signal. Active perception offers numerous advantages, enhancing both the accuracy and efficiency of the perception process. However, incorporating deep learning and active perception in robotics also comes with several challenges, e.g., the training process often requires interactive simulation environments and dictates the use of more advanced approaches, such as deep reinforcement learning, the deployment pipelines should be appropriately modified to enable control within the perception algorithms, etc. In this presentation, we will go through recent breakthroughs in deep learning that facilitate active perception across various robotics applications, focusing both on the theory that enables deep active robotic perception, as well as providing key application examples. These applications span from face recognition and pose estimation to object detection and real-time high-resolution analysis.
Next Generation of Multi-Objective Evolutionary Optimization and Decision-Making Algorithms
Sanaz Mostaghim
Otto-von-Guericke-Universität Magdeburg
Germany
![]()
Brief Bio
Sanaz Mostaghim is a full professor of computer science at the chair of Computational Intelligence and the founder and head of SwarmLab at the Faculty of Computer Science, Otto von Guericke University Magdeburg, Germany. Sanaz holds a PhD degree (2004) in electrical engineering from the University of Paderborn, Germany, has worked as a postdoctoral fellow at ETH Zurich in Switzerland and as a lecturer at Karlsruhe Institute of Technology (KIT), Germany, where she received her habilitation degree in applied computer science. Her research interests are in the area of multi-criteria optimization and decision-making, evolutionary computation, collective learning and decision-making, and their applications in robotics and science. Sanaz is a member of Saxon Academy of Sciences, the vice president of the IEEE Computational Intelligence Society (CIS), IEEE CIS distinguished lecturer, deputy chair of German Informatics and member of several advisory boards. She is an associate editor of IEEE Transaction on Evolutionary Computation as well as member of the editorial board of several international journals on AI. Sanaz has been appointed as a member of advisory board at the Ministry of Infrastructure and Digitalization, State Saxony-Anhalt, Germany.
Abstract
This talk is about the recent advances in multi-objective optimization and decision-making techniques and the future of the field. For many years such methodologies have been developed and applied to a large spectrum of applications. However, when we look at the mainstream AI, many multi-objective problems are solved using a simple weighted sum approach. A prominent example is the reinforcement learning, where only one reward function is taken into consideration. This talk is going to provide new research directions beyond the existing scope and aims to identify directions for breaking boundaries. The main focus is on the interplay between optimization and decision-making for decision support systems with focus on multi-modal optimization problems.
Assessment and Evaluation of Empirical and Scientific Data
Nikolaus Hansen
Inria & Ecole Polytechnique & Institut Polytechnique de Paris
France
Brief Bio
Nikolaus Hansen is a research director at Inria and the Institut Polytechnique de Paris, France. After studying medicine and mathematics, he received a PhD in civil engineering from the Technical University Berlin and the Habilitation in computer science from the University Paris-Sud. His main research interests are stochastic search algorithms in continuous, high-dimensional search spaces, learning and adaptation in evolutionary computation, and meaningful assessment and comparison methodologies. His research is driven by the goal to develop algorithms applicable in practice. His best-known contribution to the field of evolutionary computation is the so-called Covariance Matrix Adaptation (CMA).
Abstract
We outline some research on the assessment of performance data in optimization and reflect upon, more generally, the assessment of any empirical data or scientific information. Our considerations include methods of generic displays, in particular empirical cumulative distributions as a powerful generic tool. We argue that any ranking of algorithms should be avoided and replaced by some quantitative performance value that is comparable across publications, allows to determine effect size and has a natural zero. We also discuss the question of replicability and replication and recent concerns about the use of p-values. We finally revisit some basic methodology and caveats for how to turn data and observations into information and knowledge.