Prof. Dr.-Ing. Habil. Dr. h.c. Herwig Unger
FernUniversität in Hagen, Germany
A brain-inspired approach to sequence learning
Abstract
This talk offers insights into cutting-edge research at the Chai of Communication Networks at FernUniversität in Hagen, focusing on advancements in brain-inspired methods and the development of GraphLearner, a neuromorphic sequence analyser and generator. Beginning with a review of previous research, including text-representing centroids and a decentralized search engine, we highlight the significance of brain-inspired approaches despite the success of conventional deep learning techniques.
Drawing from the works of Hawkins and others, we introduce the GraphLearner as a promising alternative to traditional “black box” models in deep neural networks. By employing Markov models and leveraging Bloom filters for efficient probability calculations, GraphLearner, offers enhanced explainability and adaptability. The talk delves into the architecture and functioning of GraphLearner, showcasing its performance in natural language processing tasks. Additionally, we discuss potential extensions and applications of GraphLearner in diverse domains.
Biography
His research interests are in decentralised systems and self-organization, natural language processing, Big Data as well as large scale simulations. He has published more than 150 publications in refereed journals and conferences, published or edited more than 30 books and gave over 35 invited talks and lectures in 12 countries. Beside various industrial cooperations, e.g. with Airbus Industries, he has been a guest researcher/professor at the ICSI Berkely, University of Leipzig, Université de Montréal (Canada), Universidad de Guadalajara (Mexico) and the King Mongkut’s University of Technology North Bangkok.