Vloeberghs Chair 2009-1010
Prof. Dr. Philippe De Wilde

Department of Computer Science, Heriot-Watt University, Edinburgh
Philippe De Wilde

Short biography

Philippe De Wilde obtained the PhD degree in mathematical physics and the MSc degree in computer science in 1985. He was Lecturer and Senior Lecturer in the Department of Electrical Engineering, Imperial College London, between 1989 and 2005. He is currently a Professor in the Intelligent Systems Lab of the Department of Computer Science, School of Mathematical and Computer Sciences, Heriot-Watt University, Edinburgh, United Kingdom. He is also Head of the School of Mathematical and Computer Sciences.

Associate Editor, IEEE Transactions on Systems, Man, and Cybernetics, Part B, Cybernetics. Member of the Editorial Board, Intelligent Decision Technologies. Research Fellow, British Telecom, 1994. Laureate, Royal Academy of Sciences, Letters and Fine Arts of Belgium, 1988. He has published 38 journal papers and 44 conference papers and book chapters. He has published four books, including "Neural Network Models", Springer 1997, and "Convergence and Knowledge-processing in Multi-agent Syste", Springer 2

He develops biological and sociological principles that improve the design of decision making and of networks. Research interests: decision making under uncertainty; networked populations; coordination mechanisms for populations; neural networks; neuro-economics; stability, scalability and evolution of multi-agent systems.

Prof. De Wilde is a Senior Member of IEEE, Member of the IEEE Computational Intelligence Society and Systems, Man and Cybernetics Society, and the British Computer Society.

Website Prof. De Wilde

Decision making under uncertainty

Scientists need to figure out more clearly how the mind works, and then build a computer from scratch to mimic it

Gordon Moore, 2005.

One can argue where the mind is, but we all know where our brain is. Our brain is much better at making decisions under uncertainty than any existing computer. If we want to build a computer that mimics how the brain deals with uncertainty, we have to lay down a whole research programme. This programme will be introduced in six lectures.

First we have to look at historic aspects of decision making under uncertainty. It took a long time for humankind to relativize its search for certainty, and start researching uncertainty.

Mathematical models of uncertainty are dominant in statistics, machine learning and computational neuroscience. We have to understand the pros and cons of those models.

The brain is robust against damage. Some of this robustness is due to the duplication of information in neuronal populations, but an important role is played by the networks that support the neurons: the glial cells, and the cerebrovascular system of arteries and capillaries. We need models of how the astrocytes (a kind of glial cell) and the cerebrovascular system are coupled to the neural networks. An understanding of the dynamics of these interacting networks is leading to more robust learning algorithms. It is also the fundamental mechanism that allows the brain to deal with uncertainty.

Game theory and neuroimaging have come together in the new field of neuroeconomics. This provides experimentally verifiable results on decision making under uncertainty.

Bringing together mathematics, game theory, and neuroscience gives us a powerful new paradigm to model decision making under uncertainty. It brings us a step closer to artificial intelligence.

The lectures will be of interest to computer scientists, economists, actuaries, MBA, neuroscientists, psychologists and philosophers.

Six lectures on decision making under uncertainty

Participation is free of charge.

History of decision making under uncertainty (inaugural oration)

Monday 15 February 15h00-17h00, Room D2.01 (Promotiezaal)
Logic has been a help to humankind for rational decision making, but has been less useful in dealing with uncertainty. I will give a historical survey, starting with Plato, via Descartes, Russell and Wittgenstein. Most of the lecture will concentrate on the last 50 years, with probability, fuzzy logic and Bayesian inference. We will also look at subjective probabilities, and approximate probabilities.
Download presentation

An introduction to fuzzy logic

Tuesday 16 February 11h00-13h00, Room D2.01 (Promotiezaal)
Fuzzy logic is an intuitive alternative to probability theory. We will look at the fundamental operations, fuzzy arithmetic, and decision making under fuzziness. We introduce fuzzy graphs, following closely the work of the inventor of fuzzy logic, Lotfi Zadeh. Finally, we discuss Mamdani fuzzy control and it range of successful applications.
Download presentation

Coupled networks in the brain

Monday 22 March 16h00-18h00, Room E0.07
All human decision making is a result of activity in the brain. For over a hundred years, brain activity has been interpreted uniquely as activity of neurons. I will introduce a new model that takes all brain activity into account. This includes three signalling networks: neural networks, astrocyte networks, and cerebrovascular (blood flow) networks. These three networks are coupled, and both inhibit and activate each other. All our actions are a result of those processes.

Financial markets and fuzzy games

Tuesday 23 March 11h00-13h00, Room D2.01 (Promotiezaal)
Models of financial markets concentrate on rational decision makers as players. We will look at general equilibrium theory, the cornerstone of classical economics, and its implementation in agent-based computational economics. I will then introduce fuzzy decision makers, and fuzzy business games. This makes it possible to model vague information, and insider information. It is an intuitive alternative to Bayesian equilibria.

Neuroeconomics

Wednesday 24 March 16h00-18h00, Room D2.01 (Promotiezaal)
Classical economy has been adapted in the last decades to include better models of human choice. Prospect theory and behavioural economics have been tested using computer agents. The latest adaptation directly models and images brain activity. This theory of neuroeconomics incorporates emotion and value learning in social decision making. It is rewriting the old models of rational decision making. It also has practical applications in computer games.

Date, time, location

The AI of deviations of rationality

Friday 26 March 16h00-18h00, Room E0.07
In this final lecture we will draw the lessons from the previous lectures for computational intelligence. A new model of probability theory will be based on brain activity. It will be simplified, so it can be implemented as an effective computational procedure. It can be used in knowledge representation, planning, and learning. It does not assume agents are rational. I will speculate on whether it can provide the architecture for natural language processing and vision, and other processes that deviate from rationality.