Xin Yao's Research Interests: Automatic Modularisation

It is very hard to design a monolithic learning system (e.g., neural network) for a large and complex task. The divide-and-conquer strategy needs to be used in order to decompose a large and complex problem into smaller components and then solve them separately. However, decomposing a problem requires a lot of domain knowledge which may not be available or too costly to obtain for many real-world problems. I am interested in automatic decomposition of learning systems, especially using evolutionary approaches.

At present, I am interested in the design of neural network ensembles in which individual networks cooperate with each other to solve a complex problem. While trying to solve the complex problem, each individual also tries to be different from other individuals so that it will learn something that has not been learned correctly by others. A number of techniques, such as artificial speciation and negative correlation, will be studied in this project.

Selected Papers

  1. X. Yao and Y. Liu, ``Making use of population information in evolutionary artificial neural networks,'' IEEE Transactions on Systems, Man and Cybernetics, Part B: Cybernetics, 28(3):417-425, June 1998.
    Available as

  2. P. J. Darwen and X. Yao, ``Speciation as automatic categorical modularization,'' IEEE Transactions on Evolutionary Computation, 1(2):101-108, 1997.
    Available as

  3. X. Yao, Y. Liu and P. Darwen, ``How to make best use of evolutionary learning,'' Complexity International: An Electronic Journal of Complex Systems Research (ISSN 1320-0682), Vol. 3, July 1996.
    Also appeared in paper form in Complex Systems --- From Local Interactions to Global Phenomena, IOS Press, Amsterdam, pp.229--242, 1996.
    Available as

  4. P. Darwen and X. Yao (1996a), ``Automatic modularisation by speciation,'' Proc. of the Third IEEE International Conference on Evolutionary Computation (ICEC'96), Nagoya, Japan, 20-22 May 1996, pp.88-93.
    (Available as

Last update: 5 September 2000