Journal papers

Chapters

Conference papers

Convergence and Knowledge Processing in Multi-Agent Systems, Maria Chli and Philippe De Wilde, Springer, London, 2009.

'Could agent-based computer models prevent another financial crisis?', 'The economy needs agent-based modelling.' were some of the headlines in Nature on 6 August 2009. Multi-agent systems are excellent models for a range of distributed real-world problems such as trading in markets, scheduling problems, or the spread of a virus. All these real-world systems are subject to noise, and display stochastic behaviour. When they are simulated, one observes a lot of noise, and very little signal. This simulation-based analysis of complex systems has been severely criticised.

In this book, the authors develop a notion of stability from stochastic systems and apply it to multi-agent systems. They show how to use this to analyse the stability of multi-agent systems, and of the systems they model. There are special sections on business ecosystems, computer viruses, transportation of loads, diary planners, and market efficiency. A chapter is devoted to queries, and how to improve their efficiency through collaboration.

This book is just a step towards understanding convergence and knowledge processing in noisy multi-agent systems. It adds a few ideas that may just help minimise the impact of the next financial crisis...

Practical Applications of Computational Intelligence Techniques, Lakhmi Jain and Philippe De Wilde, editors, Kluwer, Boston, 2001. This book contains 12 chapters, illustrating the application of fuzzy logic, neural networks, and genetic algorithms. It starts with an introduction to computational intelligence techniques. There is a summary of fuzzy logic, backpropagation, Hopfield nets, self-organizing maps, reinforcement learning, genetic algorithms, belief networks, and computational learning theory. There is also a discussion of the synergies between neural networks, fuzzy systems, and genetic algorithms.

    


Neural Network Models, second expanded edition, Philippe De Wilde, Springer, London, 1997. This book explains and proves the main results in the dynamics of neural networks. It derives the non-linear dynamics from neurobiological principles. It investigates the stability and the convergence behavior. Another important part is the capacity of discrete neural networks. Several capacity results are derived, using signal to noise ratios, and properties of threshold functions. I also introduce stochastic networks and simulated annealing. These are studied using concepts from Markov processes, rather than statistical physics. Although I concentrate mainly on networks with feedback and their dynamical behavior, one chapter is devoted to backpropagation, a popular technique in feedforward networks.

Important parts of this book are the projects at the end of each chapter. All projects have ample introductions, and the reader's hand is held during the first part. Towards the end there are questions that could be developed into research. This book can be used for a one semester course.

Chapter 1 gives a bird's eye view of neural networks. It introduces all concepts that will be studied in depth in further chapters. The most popular neural networks are layered in structure, and they are studied in chapter 2. Neural networks are derived from models of the brain. This is the subject of Chapter 3. In this book, brain models are used as inspiration to build engineering systems, as the wings of a bird were the inspiration for the wings of an airplane. In chapter 4, we will see that neural networks are essentially dynamical systems, and we derive results about the stability of their equilibria. Chapter 5 makes the link between continuous and discrete neurons, and is also about oscillations in neural networks. The number of equilibria is calculated in chapters 6 and 7. Finally, chapter 8 introduces statistical models of neural networks. All chapters can be read independently of each other.

Journal papers

Chapters in books

Conference papers