NICAL is actively involved in advanced research topics in evolutionary computation, neural networks, pattern recognition, bioinformatics, data mining, and real world applications.

Evolutionary Computation

NICAL's research in evolutionary computation include the following major areas:

Large scale global optimisation
Most reported studies on evolutionary optimisation are obtained using low-dimensional problems, e.g., smaller than 100, which are relatively small for many real-world problems. However, most EAs suffer from the "curse of dimensionality", which implies that their performance deteriorates as the dimensionality of search space increases. So we are highly motivated to seek new evolutionary architecture for large scale global optimisation.
The focus of this research is to tackle problems that are at least one magnitude larger than the state-of-the-art in evolutionary optimisation. We do not assume the availability of any gradient information in our evolutionary algorithms.
Selected Publications:
Zhenyu Yang, Ke Tang and Xin Yao: "Differential Evolution for High-Dimensional Function Optimization", in 2007 IEEE Congress on Evolutionary Computation (CEC2007), Singapore, 2007, pp. 3523-3530.
Fundamental theories of evolutionary computation
In spite of many successes of evolutionary algorithms (EAs) in practice, the theory of evolutionary computation has lagged behind. There have been many computation studies investigating EA's performance in the literature. Our research in this area is centred around the runtime analysis of various EAs on different combinatorial optimisation problems. Computational time complexity if selected EAs have been analysed for some typical optimisation problems.
Selected Publications:
Tianshi Chen, Ke Tang, Guoliang Chen and Xin Yao: "On the Analysis of Average Time Complexity of Estimation of Distribution Algorithms", in 2007 IEEE Congress on Evolutionary Computation (CEC2007), Singapore, 2007, pp. 453-460.
Evolvable Hardware and Evolutionary Design
Evolvable hardware (EHW) refers to one particular type of hardware whose architecture/structure and functions change dynamically and autonomously in order to improve its performance in performing certain tasks. The emergence of this new field in recent years has been influenced profoundly by the progresses in reconfigurable hardware and evolutionary computation. Traditional hardware is notorious for its inflexibility. It is impossible to change the hardware structure and its functions once it is made. However, most real world problems are not fixed. They change with time. In order to deal with these problems efficiently and effectively, different hardware structures are necessary. EHW provides an ideal approach to make hardware "soft" by adapting the hardware structure to a problem dynamically.
Capacitated Arc Routing Problems
Many real-world problems, e.g., winter gritting truck routing, can be formulated as a special case of the Capacitated Arc Routing Problem (CARP). Unfortunately, finding an exact optimal solution to this problem in infeasible in practice because it is an NP-hard problem. Heuristic algorithms must be used. This research investigates hybrid EAs for CARP and its practical applications in optimising routes for winter gritting trucks.
Quantum Inspired Computation and Quantum Evolutionary Computing
The development in scientific research keeps presenting us with more and more understanding of the natural world, and consequently keeps inspiring us to develop new and more effective techniques for information processing. By now we have known that the world is quantum mechanical. As a matter of fact, quantum computation, a new research field that studies how to utilize the physical characteristics of quantum systems to implement much more efficient computation than that based on the classical computers, is absorbing more and more interests of both physicists and computer scientists.
Most emphases have been put on the study of physical realization of quantum computer and the designing of quantum algorithms that will be performed ultimately on the quantum computers. Since there are still many technical difficulties that prohibit the quantum computer to be realized right now, the benefits of quantum computation cannot be appreciated directly by peoples from other fields. While we can also take advantage of the joint of quantum physics and computer science by introducing some principles of quantum mechanics(i.e. quantum probabilistic description of state, and unitary operators to implement the state transformation) into the redesigning of some classical algorithms, such as genetic algorithm and neural network, to try to achieve some more efficient algorithms. From the stand point of view of computer scientists, it is also an obligatory and big challenge to find a way to transfer the good ideas that have been testified on classic computers to the algorithms that will execute on the future quantum computers.

Multi-Objective Evolutionary Algorithms (MOEAs)
Numerous real world problems have multiple conflicting objectives. Traditional methods based on weighted aggregation of multiple objectives have not worked well for most difficult problems. This research will study evolutionary approaches to multi-objective optimisation. Unlike most MOEAs where they only deal with a small number of objectives and decision variables, our research will focus on solving large scale multi-objective optimisation problems.

Dynamic Optimisation
So far most EA applications are for stationary problems. However, the environments of real world problems are usually dynamic, where changes occur over time. This poses serious challenges to traditional EAs since they cannot adapt well to changing environments once converged. In recent years there has been a growing interest in the research of EAs for dynamic optimisation problems because of its importance in real world applications. In this project, we propose to study novel EA approaches to dynamic optimisation problems. Both theoretical and computational studies will be cariied out.

Neural Networks and Pattern Recognition

Neural Network Ensembles
Neural network ensembles have been shown to perform better in terms of generalisation for many problems. We are interested in the issue of how to design and train a neural network ensemble so that individual networks are cooperative with each other. The idea is to have different individuals learn diferent things so that the whole ensemble can learn the overall task better. Negative correlation learning has been proposed and studied in recent years.

Kernel Learning Machines
Kernel methods are a new generation of techniques that were proposed by introducing the so-called kernel trick to traditional linear learning algorithms. Examples of kernel methods include the support vector machine, kernel principal component analysis, kernel Fisher discriminant analysis and etc. These methods have shown powerful performance on different kinds of problems, such as classification, regression, novelty detection and feature extraction. Currently, we are particularly working on kernel-based semi-supervised learning algorithms, sparse kernel machines and kernel-based feature selection methods.

Artificial Immune Systems

Artificial Immune System (AIS) is a kind of computing systems inspired by the functionalities, characteristics and theories of biological immune system. Immune System is a self-adaptive, self-learning, self-organization, parallel and distributed complex system. It is composed of many kinds of organs, molecules, lymphocytes and other cells with immune functionalities. The primary functionality of immune systme is to discriminate non-self from self, and eliminate the non-self which is harmful to the body. Immune System has many useful characteristics such as immune recognition, immune memory, immune regulation, immune tolerance, immune surveillance and so on.

The objective of this research is to extract the special information processing mechanisms contained in biological immune system, and then to design effective artificial immune algorithms for various applications, especially anomaly detection, network and Information security, optimization, evolvable hardware (EHW) and evolutionary design.