時間： 2014-09-24 15:00
報告人：Klaus Obermayer 教授
Computational Neuroscience: Quantitative Approaches Towards Reward-based Learning
Computational Neuroscience is a fairly recent field of research, where mathematical concepts and computational tools are being developed and applied to better understand brain function. In my presentation I would like to illustrate the potential of those theoretical approaches to specifically understand reward-based learning and decision making in humans.
In the first part I will discuss, how stimulus statistics and reward influence human performance in a perceptual decision making task. Combining behavioral, neurophysiological and neurocomputational approaches I will provide evidence, that early attentional mechanisms are involved in integrating this high-level information. Our results indicate, that "cognitive" quantities influence sensory processing already at its earliest stages. This hypothesis will be further investigated within the joint NPU-TUB laboratory.
In the second part I will discuss, how standard reinforcement learning can be extended to include risk, when agents face decision-making tasks in uncertain environments. In order to do so, I will present a theoretical framework, which accounts for the different "subjective" perceptions agents may have about the "objective" properties of the environment. The framework is then applied to quantify human behavior in a sequential investment task and to interpret neural activations which were simultaneously recorded by functional MRI.
Klaus Obermayer received his Diplom degree in physics in 1987 from the University of Stuttgart, Germany, and the Dr. rer. nat. degree in 1992 from the Department of Physics, Technical University of Munich, Germany.
From 1992 and 1993 he was a postdoctoral fellow at the Rockefeller University, New York, and the Salk Institute for Biological Studies, La Jolla, USA. From 1994 to 1995 he was member of the Technische Fakultaet, University of Bielefeld, Germany. He became associate professor in 1995 and full professor in 2001 at the Department of Electrical Engineering and Computer Science of the Berlin University of Technology, Germany. He is head of the Neural Information Processing Group and member of the steering committee of the Bernstein Center for Computational Neuroscience Berlin.
Since 2005 he is also guest professor at NPU. He was member of the governing board of the International Neural Network Society from 2004 - 2012 and was Vice-President of the Organisation for Computational Neuroscience from 2008-2011. From 1999-2003 he was one of the directors of the European Advanced Course of Computational Neuroscience. His current areas of research are computational neuroscience, artificial neural networks and machine learning, and the analysis of neural data. He co-authored more than 250 scientific publications.
2004 - 2012年，國際神經網絡學會理事會成員沙巴体育中国版，并在2008-2011年擔任理事會副主席