Autonomous Incubator Seminar Series:
RISK CONTROL IN THE COGNITIVE DYNAMIC SYSTEM
Simon Haykin, Cognitive Systems Laboratory, McMaster University, Ontario, Canada
September 23, 2015, 10:00 am, NASA Langley, Pearl Young Theater
Abstract:
The Cognitive Dynamic System (CDS) has established itself as the basis for a new generation of Intelligent Machines, inspired by the brain, and with varying applications: Radar, Control, and Radio. The underlying principles of this new way of thinking naturally involve Cognitive Neuroscience, and Bayesian dynamics, and to a lesser extent, Bellman’s dynamic programming.
As the title of the lecture would imply, the primary objective of this lecture is how to control risk: A topic that has been broadly discussed in the literature, but seriously lacking a systematic procedure for how to deal with it.
Here are the issues that will be discussed in the lecture:
- The perception: The Bayesian model for environmental observables, reciprocally coupled with the cubature Kalman filter for state estimation
- The feedback channel for linking the executive to the preceptor
- Executive reinforcement learning, followed by predictive pre-adaptive subsystem; it is here that risk control is realized, incorporating the MAP rule for decision-making.
The lecture will be finished by highlighting Dr. Haykin’s vision for the next five years, focused on two topics:
- The Bayesian dynamics of hierarchical CDS for reducing unexpected disturbances down to a precision level that can be ignored for all practical purposes.
- The CDS employed as the supervisor of Cognitive Dynamic Networks, aimed at network science.
Bio:
Simon Haykin received his B.Sc., Ph.D., and D.Sc. in Electrical Engineering from the University of Birmingham, England, where after he spent a good number of years in industry, then moved into the world of academe in England, and finally he joined the ECE Department, McMaster University, where he is Distinguished University Professor.
For the early years, his research program focused on adaptive signal processing for radar and communication systems, then spending more of his time on Neural Networks and Learning Machines, which in turn led him to research into the Cognitive Dynamic System with pioneering applications in radar, control, and radio. Needless to say, the new generation of Cognitive Dynamic Systems and Networks has become his passion. Professor Haykin is a Fellow of the IEEE, Fellow of the Royal Society of Canada, and the recipient of an Honorary Doctorate of Technology from ETH, Zurich, Switzerland. He is also a recipient of many medals from within Canada, the United States, and Europe, and the author of many classic books that have impacted the world of communications, adaptive systems, and neural networks. E-mail: haykin@mcmaster.ca