Milan Segedinac

University of Novi Sad, Faculty of Technical Sciences

Biography

Milan Segedinac received his M.Sc. in 2008 and Ph.D. in 2014 in Computer Science both from the Faculty of Technical sciences University of Novi Sad where he currently holds the position of a full professor in applied computer science and informatics. His research focuses on Artificial Intelligence and Software Engineering, particularly AI-driven educational technologies and knowledge representation in computer-supported education.

He has authored and co-authored over 50 research papers published in international and national journals or presented at international and national conferences, especially in the field of technology-enhanced learning. He has also participated in more than 20 commercial and scientific projects.

In addition to his academic contributions, Milan Segedinac collaborates with universities and research institutes worldwide, working on advancing AI applications in education and fostering innovation in intelligent learning systems.

Keynote speech title

Bridging Symbolic and Connectionist AI: Neuroevolutionary Approaches to Graph-Based Knowledge Representation

Abstract

Graph-based methods for knowledge representation are among the oldest techniques in AI, yet they remain crucial today, particularly in applications like knowledge graphs. At the same time, connectionist AI, based on neural networks, has seen remarkable advancements. In recent years, the boundary between symbolic and connectionist AI has become increasingly blurred, leading to the rise of neuro-symbolic AI. This convergence offers new opportunities to combine structured reasoning with learning-based approaches.

In this talk, we explore a novel method that applies neuroevolutionary techniques to the construction of knowledge spaces. Traditional approaches to building these mathematical models often struggle with scalability, making them less effective for large and complex domains. The approach that we discuss in this talk bridges the gap between connectionist and symbolic models by defining a set of analogies between knowledge spaces and neural networks, allowing evolutionary algorithms to shape an optimal structure for knowledge representation.

To illustrate the potential of this method, we focus on its application in AI-augmented learning management systems, where it helps model student learning and adapt educational pathways. By integrating neuro-evolution with graph-based knowledge representation, this approach offers a new way to enhance adaptive learning and contributes to the broader effort of bridging symbolic and connectionist AI.