Abstract

Neural Information Processing and Learning: From Biological to Artificial Systems

The human brain is an enigmatic organ. Not everyone is aware of the fact that each human carries beneath the skull an information processing device that is more powerful than a modern supercomputer while consuming many orders of magnitude less energy. The algorithms implemented in the brain endow us with intelligence and powerful learning capabilities. It is a long-standing dream of mankind to understand the principles of computation and learning in the brain and to build artifacts with similar abilities. 

In this talk, I will guide the audience through my research that is motivated by the following questions (a) which principles underlie computation and learning in the brain, and (b) how can these principles be utilized for artificial information processing systems. I will focus on theoretical and modelling work on synaptic plasticity (that is, changes in the strength of connections between neurons). Rewards are a major driving force for learning in humans and animals. I will discuss models for reward-based learning in the brain. I will also show how synaptic plasticity in cortical microcircuits can lead to the formation of assembly codes. Finally, I will present examples of how these insights can be used to improve modern deep learning methods and develop learning paradigms for novel computing hardware with nanoscale circuit elements.

Bio

Dr. Robert Legenstein received his PhD in Computer Science from Graz University of Technology in 2002. He is currently Associate Professor and the head of the Institute for Theoretical Computer Science at the Department of Computer Science, Graz University of Technology. Dr. Legenstein is especially interested in biologically inspired neural computation. His research interests include neural networks, learning in neural systems, reward-based learning, spiking neural networks, information processing in biological neural systems, and deep learning. Dr. Legenstein is associate editor of IEEE Transactions on Neural Networks and Learning Systems.