Distinguished Lecture: Cognitive Risk Control for Physical Systems, presented by Dr. Simon Haykin

Friday, February 17, 2017 1:30 pm - 3:00 pm EST (GMT -05:00)

Abstract:

The Cognitive Dynamic System (CDS) is a unique research tool inspired by certain aspects of the brain; one aspect is "predictive adaptation," known in cognitive neuroscience. From an engineering perspective, predictive adaptation plays a key role in the CDS, most importantly when the environment is disturbed by the presence of unexpected adverse events, commonly referred to as ‘risk.’

The lecture will address how Cognitive Risk Control in the CDS has information-processing capacity to tackle the challenging problem of risk. The presentation will begin with a discussion of the first principle of cognition: the "perception-action cycle,” followed by three additional principles – memory, attention, and intelligence. "Perception" of the environment (representing physical systems) will be examined by appealing to the "Bayesian choice", then finishing with the "entropic state" of the perceptor.

The next functional block of the CDS, "feedback channel,” links the executive to the perceptor using “internal rewards.” As for the remaining functional block, the "executive", will be dealt with in two parts. For Part A, "reinforcement learning" is followed by "cognitive control," which are integrated together for "cognitive action" on the environment in the absence of uncertainties. It is here where "executive memory" comes into play to accumulate past experiences (i.e., past cognitive actions). As for Part B of the executive, the issue is more sophisticated because it operates in the presence of uncertainties:

  1. cognitive control results in "perturbed cognitive action;”
  2. at the same time, the second feedback channel called "perceptual curve" activates the executive memory for presenting a "prospective set of past experiences;" and
  3. the "classifier" matches the perturbed cognitive action against the prospective set of past experiences. In so doing, it selects from the executive memory the desired "cognitive risk sensitive action.”

An important point to note: "switching mechanisms" integrate points (1), (2), and (3) to account for the executive as the most important functional block of the CDS.

Briefly described here is a "brand new way of thinking" for bringing risk under control for the first time ever.

Biography

Dr. Haykin received his B.Sc., Ph.D., and D.Sc., all in electrical engineering, from the University of Birmingham, England. After graduating, he worked at GEC (Telecommunications) in Coventry, England before returning to academic life at Lanchester College of Technology and then the University of Warwick. Dr. Haykin joined McMaster as a full professor in the Department of Electrical and Computer Engineering in 1966. In 1992, McMaster University honoured Dr. Haykin with the title Distinguished University Professor.

Dr. Haykin established the first Communications Research Laboratory (CRL) in Canada, which was widely recognized for its contributions to signal processing, communications, and radar applications.  He is among the most cited engineering researchers in the world. He has authored or co-authored close to 50 textbooks and numerous academic papers in leading journals; his two most cited publications are: Neural Networks: A Comprehensive Foundation, first published in 1994, garnering more than 18,000 citations in academic articles: and Adaptive Filter Theory, which has generated more than 12,000 citations. Both are used as advanced university textbooks on the fundamentals of communications and radar systems. The use of Dr. Haykin’s textbooks is so widespread that many, if not most, of today’s practicing engineers learned their fundamentals in communications, radio, and radar from him. Dr. Haykin’s current research focus is Cognitive Dynamic System Theory.

Dr. Haykin is a Fellow of the Royal Society of Canada, Life Fellow of the Institute of Electrical and Electronics Engineers (IEEE), and recipient of many international awards, medals and prizes. In 2016, Haykin was awarded the prestigious IEEE James H. Mulligan, Jr. Education Medal.