The world around us is a complex and dynamic system. As a result, we need intelligent systems that can learn and adapt in a changing environment without human intervention in order to solve complex and dynamic real-world problems. Cercia is an international leader on these topics and has in-house expertise in solving real world problems using these techniques for many years.

Warehouse Task Scheduling

Cercia investigated algorithms that can prioritise task lists and allocate individual tasks to vehicles and personnel. For efficiency, the solutions should distribute new work items on an on-demand basis, taking into account all the relevant constraints. This process needs to be extremely rapid, with the interval between independent task completions potentially measured in seconds. Individual staff should receive a new task instantaneously, when requested. The work distribution should be adaptive, able to cope with warehouse configuration changes such as incorrectly-recorded (or damaged) stock; new personnel arriving for work, or existing personnel leaving; vehicles breaking down; job prioritisation changes; new incoming stock; new purchase orders; updates to the warehouse state.

(supported by Severn Trent)

Financial planning and forecasting are essential for any business. Turning "gut feeling" into sound decision-making based on real data, is a huge challenge for both large corporations and SMEs, as well informed decisions mean profits. Severn Trent Water Ltd. has recently turned to the Centre of Excellence for Research in Computational Intelligence and Applications (Cercia) at the University of Birmingham for state-of-the-art solutions to emerging business challenges. Two phases of a software project in financial planning and forecasting were completed in early 2003 and 2005, respectively. This software has been in use by Severn Trent Water since early 2003 and a major upgrade was completed in early 2005.

Decision Aid for Medical Diagnosis

Early diagnosis of many diseases is crucial to the successful treatment of an illness. However, it is often very difficult to diagnose a disease correctly based on symptoms and test results. We have developed a very successful evolutionary artificial neural network system, i.e., EPNet, for assisting a doctor to diagnose diseases, including heart disease, diabetes, breast cancer, and thyroid [1,2,3]. The system was tested on real-world data sets and compared against other similar systems [1,2,3]. The results showed that our system performed significantly better than others [1,2,3].

Although we have only applied our system to four types of diseases, it can be applied to other diseases. The key idea behind our system is to use a special evolutionary programming algorithm to evolve both the architecture and weights of the artificial neural network automatically [1,2,3].

The same principle and techniques can also be adapted to the diagnosis of other faulty systems, e.g., engines, manufacturing systems, etc.

Environmental Monitoring and Protection

Environmental monitoring and protection are crucial to the quality of our life and the ecological system in the world. For example, a major outbreak of the blue-green algae in a fresh water lake can have a devastating impact on the fish and other lifes in the lake.

We have used evolutionary artificial neural networks to predict the activities of Chlorophyll-a in a lake in Japan [4]. We were able to produce much more accurate prediction than other methods [4]. The more accurate prediction would enable us to take an appropriate preventive action and reduce the risk of a major outbreak.

Traffic Flow Prediction in Telecommunications Networks

In a telecommunications network, it is very important to know when the network is busy and when it is not. Such information will enable a carrier (company) to make an informed decision on the necessary capacity between two cities as well as set a pricing policy that encourages off-peak use of certain lines. However, it is very difficult to predict traffic flow in a telecommunications network. We have used our newly developed neural network ensembles to predict traffic flow in an Austrian telecommunications network among 32 regions [5]. We have shown that our negative correlation learning algorithm can train an ensemble successfully to solve this problem.

Although the project was carried out for a telecommunications network, the techniques we used can equally be applied to flow prediction in an electricity network, a water network, or a gas network. We will be able to predict the demand within a period based on historical data.

Credit Card Assessment

Credit cards are widely used in the world, even in developing countries like China and India. However, issuing a credit card carries some risk because the card holder may or may not pay back the money. It is important for any bank to assess the risk of a potential card holder before issuing a card. We have applied our evolutionary artificial neural networks and neural network ensembles for credit card application assessment in an Australian bank [2,6,7]. Excellent results have been obtained in comparison with other existing methods.

The techniques we have developed for the Australian credit card problem can be applied to other problems, such as insurance fraud detection [8], risk assessment for loans, premium setting for insurance, etc.

Financial Time-Series Analysis and Prediction

The ability to analyse and predict the movement of financial time series is crucial to the success of any financial institution. We have applied our evolutionary artificial neural networks to the analysis of Hang Seng Stock Index and obtained impressive results [9]. More work can be done on portfolio optimisation, bond price prediction, share price analysis, etc.

Object Recognition and Automatic Feature Extraction

Many real world problem involves object/pattern recognition, e.g., automatic inspection of products through photos, videos or infrared, target recognition in an image, finger print recognition, number plate recognition, etc. This kind of problems usually requires a recognition system that can deal with a large number of inputs. However, not all inputs may be important to the recognition task. It is essential to extract only relevant features out of a vast number of possible inputs so that a high accuracy recognition system can be constructed. We have invented (EU patent pending) a novel method for automatic feature extraction and object recognition based on neural network ensembles [10], which can be applied to a wide range of real-world recognition problems [11].

Adaptable Sensor/Motor/Robot Control Systems

Many manufacturing and processing systems involve sensors, motors and robots that must be controlled to operate in a smooth and efficient manner. Such controllers are frequently required to adapt to changing environmental conditions, plant wear, and such like. Rather than attempting to design these systems by hand, one can set up sufficiently general neural network or traditional control systems, and let them learn from appropriate training data sets how best to operate. Such systems can be trained to perform well according to a range of different performance criteria [19]. Even more powerful systems can be developed by allowing whole populations of these control systems to evolve by a process of simulated natural selection [20, 21].


A common real world problem is the need to recognise patterns in data, and use them as a basis for classification. Standard neural network systems are well known to be able to do this well. More specialist neural network systems can be tailored to cope with ambiguous training data [22], and various types of noisy training data [23]. Further insights into the data and the tasks under consideration can be obtained by analysing the internal representations learnt by the neural networks [24]. Complex multi-task problems, such as the requirement to perform two or more distinct classifications at once, may benefit from building some form of modularity into the system, and evolutionary strategies can be employed to optimise these more complex architectures [25]. Similar evolutionary approaches can also be used to optimise the speed of learning in neural network systems [26], which will be particularly important when real-time training is required.

Design and Optimisation

Design and optimisation are ubiquitous, from circuit design to toy design, from turbine blade design to truss design, from inventory minimisation for a company to wastage minimisation in stock cutting, etc. Computational intelligence techniques have proven to be extremely competitive in solving complex real-world problems in comparison with existing mathematical programming methods. Cercia is at the forefont of applying computational intelligence techniques in real world design and optimisation problems.

Nature Inspired Creative Design Network

"Nature Inspired Creative Design" is a new research network, part of the AHRB/EPSRC funded "Designing for the 21st Century" program. Cercia is creating a cluster of scientists, artists, designers, and industrialists that focusses on taking ideas, methods, paradigms and algorithms from nature and introducing them into the design process. Nature inspired approaches have the potential to create better designs, as well as better design processes and tools.

Stock Cutting

Stock cutting is a problem faced by many workers in the world. For example, when steel pipes are manufactured, they are produced in certain fixed length. However, customers normally have different requests for pipes of different length. To save costs, we need to minimise wasted remainders. We have invented a novel evolutionary programming algorithm that can cut stocks efficiently by minimising the wastage [12]. The algorithm performed significantly better than other algorithms on a number of problems we have tested [12]. The algorithm can be used to cut any one dimensional stocks.

Container Packing

Container packing could be regarded as a 3-d version of stock cutting. The problem occurs daily in transportation and logistics companies. The key issue here is: How can we pack as many goods (of different sizes) as possible to large and fixed-size containers so that we use as few containers as possible? We have developed a software package on PC with a user-friendly GUI to perform container packing automatically. A user can enter the size of containers and sizes of the goods to be packed. With button-clicking, the software will display the packing that it generates using 3-d graphics.

Material Modelling

Designing new materials, such as alloys, with certain properties is always a difficult and tedious task because we do not know all the properties of the material until we physically produce and test it. Material modelling is extremely important in understand various properties of the material. We have teamed up with a material scientist and developed a novel evolutionary approach to material modelling [13,14]. Our approach has been able to obtain the results that no other methods could. We have also developed a software package for material modelling. The software is generic and not limited to any particular type of materials.

Discovery of Novel Digital Filters

Evolutionary algorithms are very good at discovering novel designs that can hardly be hit upon by a human designer. We have invented (UK and EU patent pending) a new method based on multiple Pareto fronts to evolve various designs [15,16]. The advantages of our method have been illustrated by a novel digital filter that we evolved [15,16]. It is worth noting that our method is not limited to digital filter design. It can be applied to other design tasks.

Terrain Traversal

Finding the shortest as well as smoothest route between two locations is extremely important in designing new roads or pipelines for gas, water and oil. The problem is often complicated by the constraints, such as natural reserves, residential areas, archaeological sites, etc. We have developed an evolutionary algorithm based system for finding the shortest as well as smoothest route between any two locations in a map [17]. Comparisons with human designers have shown that our evolutionary system can find better solutions than human beings. The system can be used for many design as well as planning tasks.


Generating a timetable for a university or a sport event (e.g., the Olympic Games) can be a nightmare for the administrators because the number of courses/events and constraints involved. We have developed a timetabling system for the Australian Defence Force Academy [18], which produced results that were better than those generated by the timetabling secretary. Our system is also flexible and can be adapted to other requirements from a different organisation, e.g., another university or sporting event.

Computational Intelligence in Manufacturing

Computational Intelligence (CI) finds strong applications in manufacturing. CI techniques help us solve complicated problems for which no satisfactory solutions exist. These applications include:

  • Manufacturing process monitoring
  • Machining error analysis and compensation
  • Defect detection in castings
  • Stamping process monitoring

A few examples of computational intelligence in manufacturing are presented here.

Tool Condition Monitoring And Machining Error Compensation In Metal Cutting

Real-time monitoring of modern CNC machining equipment is hard but also essential to prevent damage to both machine tools and expensive work pieces. We have used a combination of Acoustic emission (AE), Cutting Force, Motor Current and Vibration Sensors to monitor real-time performance and accuracy of machine tools. We use a variety of CI techniques to analyse the data captured by these sensors and real-time control based on this information is able to improve the accuracy of the controlled machine considerably without a significant increase in investment.

Detection of Defects in Casting

X-ray based inspection systems are widely used for the identification and evaluation of internal defects, such as cracks, porosities and foreign inclusions in castings. We have used Wavelet Transforms and Imaging Processing techniques to detect internal defects using X-ray images and an improvement over traditional techniques.

Monitoring Sheet Metal Stamping

Sheet metal stamping is an extremely popular manufacturing process. In-process detection of the malfunctions and on-line monitoring can help ensure product quality and also help protect expensive dies and presses. This is particularly important for unmanned modern stamping operations.


ChessBrain - The World's Largest Distributed Chess Network Researchers

Colin Frayn, Carlos Justiniano (Private researcher, USA) Summary

The ChessBrain project was created to investigate the feasibility of massively distributed, inhomogeneous, speed-critical computation on the Internet. The game of chess lends itself extremely well to such an experiment by virtue of the innately parallel nature of game tree analysis. We believe that ChessBrain is the first project of its kind to address and solve many of the challenges posed by stringent time limits in distributed calculations. These challenges include ensuring adequate security against organized attacks; dealing with non-simultaneity and network lag; result verification; sharing of common information; optimizing redundancy and intelligent work distribution. [1,2,3,4]

ChessBrain was formed by private researcher Carlos Justiniano, based in Los Angeles, US, in early 2002. Colin Frayn (Cercia) joined the project a few months later, and added his master-level chess engine, 'Beowulf'. He also worked on the fundamental communication strategy between the chess-playing code and the networking infrastructure. His considerable input centred on the protocols necessary to allow a chess engine to split up the available work into manageable chunks and to distribute them sensibly to connected machines over the internet.

The project culminated with a demonstration match on 31st January 2004 against Denmark's top Grandmaster player, Peter Heine Nielsen. This match resulted in a draw after 34 moves. A total of 2,070 independent users contributed to the event, which was acknowledged as a world record by the Guinness organisation on 13th February 2004. It will appear in the 2005 edition of the book under "Largest networked chess computer". External Funding

Private funding, grants from Y3K Secure Enterprise Software Inc. and the Danish UNIX Users Group (DKUUG). Selected References

[1] Justiniano, C., & Frayn, C.M., The ChessBrain Project, 2003 ICGA Journal, Vol. 26, No. 2, 132-138
[2] Justiniano, C., ChessBrain: a Linux-Based Distributed Computing Experiment, Linux Journal, September 2003
[3] Justiniano, C., Tapping the Matrix, O'Reilly Network Technical Articles, 16th, 23rd April 2004.
[4] Frayn, C.M. & Justiniano, C., The ChessBrain project - Massively distributed, speed-critical computation, Proceedings IC-SEC Workshop on Grid Computing and Applications, 2004, 10-13

Force Based Clustering and Visualisation Researchers

Colin Frayn, Andy Pryke Summary

Cercia has developed a considerable expertise in the field of force-based visualisation. Following on from the results of [1], we decided to implement a more advanced 3D visualisation suite based on established techniques, but extended with the use of modern hardware and software solutions.

The many physical forces that shape our world inspire force-based visualisation. Just as nature creates organised structures in the natural world, physically inspired visualisation techniques allow us to extract potentially valuable information from complex data. Force-based data clustering works by randomly placing sets of data points as particles into a three-dimensional space, and then joining them together with forces related to their similarity. Very similar points are joined with strong forces, whereas weaker forces join those that are not so similar. All particle pairs have an ideal mutual separation that they are aiming to reach, keeping similar particles close and dissimilar particles further away. As the system relaxes, the particles form into clusters based on their common properties.

The result of this is a self-organising system that automatically seeks out its most efficient configuration. An advanced 3D visualisation environment allows any user to explore the structure in an intuitive and interactive manner.

The applications of this technology are very far-reaching. Cercia is currently working with W3 Insights [2], an educational psychology company, to integrate our technology into their existing methodology. We are also working on applying this technology to Company template-matching [3], Internet server log analysis and visual network analysis. We believe that it offers a considerable improvement on the current state-of-the-art in all of these areas. Selected References

[1] Pryke. A, (1999). Data Mining using Genetic Algorithms and Interactive Visualisation (PhD Thesis), The University of Birmingham.
[2] Pryke. A, Cercia Business Assist Report 2004/05-I-2, 2004
[3] Frayn. C, Cercia Business Assist Report 2003/04-I-4, 2004

Modelling Radial Brightness Distributions in Elliptical Galaxies Researchers

Jin Li, Xin Yao, Colin Frayn, Habib G. Khosroshahi (Physics and Astronomy, University of Birmingham), Somak Raychaudhury (Physics and Astronomy, University of Birmingham) Summary

Modelling the galaxy surface brightness as a function of radial distance from its centre is aimed at quantifying the morphological features in an attempt to link them to the physics of their formation, whilst developing a deeper understanding of the evolution of galaxies.

To obtain a radial brightness profile model of a galaxy, the method varies based on the exact mathematical form of the function used, and the algorithm used for parameter fitting. The traditional approach takes two steps: firstly, it assumes a mathematical model in advance; and secondly, it applies fitting algorithms to find suitable parameters for the assumed function. The parameter-fitting algorithm usually adopted is the non-linear reduced ?2 minimization.

There are several drawbacks of this approach. In general, it is hard to select appropriate functional forms. The form selected is ad hoc in that it may merely suits a smaller number of profiles. Moreover, the parameter-fitting algorithm is prone to getting trapped in local optima.

We proposed a novel approach that enables one to build profile models from data directly without assuming a functional form in advance by using CI techniques. The novel evolutionary approach takes two steps as well. The first step is to apply the genetic programming (GP) technique to find functional forms; the second step is to apply the evolutionary programming (EP) technique to carry out parameter fitting. The proposed evolutionary approach overcomes many of the drawbacks of traditional methods. We have tested on a set of 18 elliptical galaxies in the Coma cluster observed in near-infrared band are used. Experimental results demonstrate that it is superior to the traditional approach with the following advantages:

It is a data-driven process without assuming a functional form beforehand
It is a bottom up process that potentially suits modelling a large number of galaxy profiles without any prior knowledge
It generates explicit formulae enabling interpretation of physical meanings
It is more effective in parameter fitting

The methodologies in this work could be potentially applied to other fields where modelling is desirable even if no prior knowledge is available. Examples of other fields include carbohydrate and protein chemistry, predictive modelling techniques in materials science and materials engineering. External Funding

This work has led to a successful PPARC grant worth �152,000 for designing new algorithms for data mining in astronomy. Selected References

Li. J, Yao. X, Frayn. C, Khosroshahi. H.G, and Raychaudhury.S, (2004). An Evolutionary Approach to Modeling Radial Brightness Distributions in Elliptical Galaxies, to appear in the Proceedings of the 8th International Conference on Parallel Problem Solving from Nature (PPSN VIII)

Evolvable Hardware and Automatic Discovery of Unconventional Designs Researchers

Thorsten Schnier, Xin Yao Summary

The major area of the research is the development of an evolutionary software system that enables us to experiment with various ideas we have on evolvable hardware and unconventional design. This provides a much cheaper alternative to experiment with hardware directly. It has been shown to be sufficient in proving the concept and feasibility of our ideas. Our research includes several new techniques that can be used in evolvable hardware and unconventional design. Extensive computational studies have been carried out to evaluate the strength and weakness of the new techniques, e.g., cluster Pareto optimisation with multiple fronts, adaptive fitness sharing, self-adaptive constraint handling, etc. Three papers [1, 2, 3] have been published and one patent application [4] filed out of the work from this project. We have also published annual reports and a final report [5].

This initially three-year project unfortunately had to be terminated after two years due to changed priorities of the project sponsor. However, we have made major progresses in all three areas that were set out in the initial three-year plan, i.e., evolving novel designs, discovery of useful genes in an individual and maintaining diversity. External Funding

Marconi PLC. Selected References

[1] Schnier. T, Yao. X, and Liu. P, "Digital Filter Design Using Multiple Pareto Fronts," Proceedings of the Third NASA/DoD Workshop on Evolvable Hardware, pp.136-145, IEEE Computer Society Press, July 2001.
[2] Schnier. T, and Yao. X, "Evolutionary Design Calibration," Proceedings of the 4th International Conference on Evolvable Systems (ICES-2001), LNCS 2210, pp.26-37, Tokyo, Japan, October 3-5, 2001.
[3] Schnier. T, Yao. X and Liu. P, "Digital filter design using multiple pareto fronts," Soft Computing, 8(5): 332-343, April 2004
[4] Yao. X, Schnier. T, "Hardware Design Using Evolutionary Algorithms", Patent Application, PCT WO02075650, 2001
[5] Yao. X, Schnier. T, "Evolvable Hardware and Automatic Discovery of Unconventional Designs", Final Project Report, 2002

Fault-Tolerant Hardware through Artificial Evolution Researchers

Thorsten Schnier, Xin Yao Summary

Diversity is of key importance in deploying redundancy for fault-tolerant circuits - without it, single faults can cause redundant circuits to fail with common failure modes. Creating diverse circuits using conventional means is difficult, as these methods can usually only produce very few variations for any particular design task. Evolutionary algorithms, however, are not restricted by the available design methods, and search a much larger space. Apart from circuits that may be faster and/or smaller, they are generally also very different; both compared to conventional circuits, and to ones created in separate runs.

Our approach expands on this by incorporating fault tolerance into the fitness, thereby evolving sets of circuits that together have optimal fault tolerance. Selected References

Schnier. T, and Yao. X, "Using negative correlation to evolve fault-tolerant circuits," In Proc. of the 5th International Conference on Evolvable Systems: From Biology to Hardware (ICES'2003), Lecture Notes in Computer Science, Vol. 2606, A. M. Tyrell, P. C. Haddow and J. Torresen (Eds.), Springer-Verlag, March 2003, pp.35-46.

Intelligent Energy Consumption Monitoring: Research and Software Development Researchers

Thorsten Schnier Summary

Remote metering is increasingly being used to collect data from water, electricity, heat, and gas consumption meters. Cercia has developed an intelligent monitoring system that allows site owners to automatically monitor meter readings. The tool enables users to spot unusual activity, thereby creating opportunities for cost savings External Funding

We have applied for a Mercia Spinner investment grant, which is under review. Traffic Grooming in WDM Optical Networks Researchers

Yong Xu, Xin Yao, K. Liu (Fujian Normal University), S. Xu and B. Wu (Xiamen University). Summary

Traffic grooming is a traffic engineering technique aimed at achieving efficient utilisation of network resources, e.g., minimising cost or blocking probability, maximising revenue by assigning sub-wavelength traffic flows onto wavelengths. This is a complicated combinatorial optimisation problem with high academic and commercial values and has been one of the most attractive issues in optical network areas. The objectives of this project are to conduct theoretical and practical research of grooming problems in Synchronous Optical Networks (SONET) / Wavelength Division Multiplexing (WDM) ring networks, especially by using metaheuristic approaches.

Cercia first investigated static grooming problems in SONET/WDM rings with a GA approach [1]. In this paper, we proposed a GA approach to directly assigning traffic flows onto wavelengths in a synthesised way. We then dealt with the grooming of dynamic traffic in such networks with both splitting and non-splitting methods [2-4]. In [2], we proposed a comprehensive GA approach to carrying out strictly non-blocking grooming of dynamic traffic without splitting. We treated the M traffic patterns as a whole and incorporated a GA approach with a local search mechanism to achieve better results. In [3], we analysed the benefit of grooming by splitting and combined the GA approach with splitting techniques to further improve grooming results. In [4], we incorporated the splitting method with a GA having hierarchical chromosome structure to tackle the rearrangeably nonblocking grooming problems. In this algorithm, each individual is composed of two hierarchical chromosomes: a master chromosome and M slave chromosomes for a ring with M traffic demands. The master chromosome is used to record each wavelength's dropping nodes and M slave chromosomes to track the assignments of M traffic demands among wavelengths. Furthermore, the slave chromosomes are only accompanying chromosomes which are produced by a heuristic to rearrange traffic demands among wavelengths and then delete some spare ADMs and/or wavelengths after such rearrangement. The slave chromosomes are updated when a new master chromosome was generated by crossover or mutation. Our theoretical research and the use of tabu search approach to the problem can be found in [5]. In this paper, we derived a lower bound on the number of ADM's with arbitrary traffic demands in ring networks, which is the tightest by now. Finally, we presented a comprehensive, thorough, and up-to-date review of the metaheuristic approaches to the grooming of both static and dynamic traffic in WDM optical networks in [6]. Future directions concerning the application of the metaheuristic approach in the grooming of traffic in various aspects are also discussed in this paper.

In the future, what we are going to do is the dynamic grooming of traffic in GMPLS-based IP/WDM networks. Other topics of interest are multi-cast traffic grooming and grooming with QoS considerations etc. On the other hand, the techniques and algorithms used in this project can be directly applied to many other similar optimisation problems, such as bin-packing problem, vehicle routing problem, transportation problem, electric power, water, or gas supply problem, etc., all of which are the potential targets of our future research. Selected References

[1] Xu. Y, Xu. S. C, and Wu. B. X, Traffic grooming in unidirectional WDM ring networks using genetic algorithms. Computer Communications, 25(13): pp.1185-1194, 2002.
[2] Xu. Y, Xu. S. C, and Wu. B. X, Strictly nonblocking grooming of dynamic traffic in unidirectional SONET/WDM rings using genetic algorithm. Computer Network, 41(2): pp.227-245, 2003.
[3]Liu. K, and Xu. Y, A new approach to improving the grooming performance with dynamic traffic in SONET rings. Computer Networks, 2004 (to appear).
[4]Xu. Y, and Yao. X, Grooming of dynamic traffic in SONET rings: MOINLP formulation and tabu search approach submitted to IEEE/ACM Trans. on Networking, Oct 2003.
[5]Xu. Y, and Yao. X, Lower bound on number of ADM's in WDM rings with non-uniform traffic demands. Electronics Letters, 40(13), pp. 824-825, June 2004.
[6]Xu. Y, Salcedo-Sanz S, and Yao. X, Metaheuristic approaches to traffic grooming in WDM optical networks", International Journal of Computational Intelligence and Applications, 2005 (to appear).

Terminal Assignment in Telecommunications Networks Researchers

Sancho Salcedo-Sanz, Xin Yao, Yong Xu Summary

Terminal assignment (TA) is an important issue in telecommunication networks' optimisation to increase their capacity and reducing the cost of them. It is a NP-complete combinatorial optimization problem in which terminals having a known requirement of capacity have to be assigned to a given concentrator with a given maximum capacity. The objective of the TA is to minimize link cost to form a network by connecting a given set of terminals to a given collection of concentrators.

Cercia has investigated the terminal assignment (TA) problem in a telecommunication network with a novel hybrid Hopfield network (HNN) GA approach and a tabu search approach [1, 2]. In [1], we focused on TA instances where only the cost function of entire feasible solutions can be calculated and the cost of a single assignment cannot be known in advance. We present a novel hybrid Hopfield network (HNN) GA in which the problem's constraints are managed by the HNN and the quality of the solution obtained is improved by the GA. The use of the HNN reduces the search space of the GA to only the feasible solutions. The performance of our hybrid algorithm is evaluated in several test TA problems with very good results in all test instances considered. In [2] we proposed a two-objective model of the TA problem. The two objectives considered in this research are the cost of each terminal assigned to a concentrator and the balance of terminals distributed among concentrators. We present a tabu search approach to solve two extremes of this TA problem, i.e., the cost-first and the balance-first cases, respectively. Simulations showed that the minimal cost and the best balance of this problem can not be reached simultaneously. When the balance-first approach is applied, the cost is 10 to 75 percent higher than the cost-first approach. If the cost-first approach is used, the balance is about 20 to 50 percent higher than the balance-first approach. But the other objective could always reach its optimum. Experiments showed that the comprehensive performance of our approach is better than the previous ones.

This hybrid Hopfield network (HNN) GA approach has also been applied to other problems directly related to TA, i.e., the task assignment problem [83] and FPGA segmented channel routing problems (FSCRPs) [84], in order to show the effectiveness of this approach to other problems. We showed that our algorithm is able to obtain very good results for these problems, outperforming the others GAs within a reasonable computation time. Selected References

Salcedo-Sanz S, Yao X. A hybrid Hopfield network-genetic algorithm approach for the terminal assignment problem. IEEE Trans. System, Man and Cybernetics B, 2004, 34(6): 2343-2353.
Xu Y, Salcedo-Sanz S, Yao X. Non-standard Cost Terminal Assignment Problems Using Tabu Search Approach. Proc. of the 2004 Congress on Evolutionary Computation (CEC'04), IEEE Press, Portland, Oregon, USA, 19-23 June 2004, pp.2335-2340.
Salcedo-Sanz S, Xu Y, Yao X. A Hybrid Genetic Algorithm-Hopfield Network for Task Assignment in Distributed Computer Networks. Computers and Operations Research, 2005 (to appear).
Salcedo-Sanz S, Xu Y, Yao X. Meta-heuristic Algorithms for FPGA Segmented Channel Routing Problems with Non-standard Cost Functions. Genetic Programming and Evolvable Machines, Vol. 6, 2005 (to appear).

Target Classification and Recognition Using Neural Network Ensembles Researchers

Xin Yao, Peter Duell (PhD student), Sophie Kain (Thales), Yong Liu (Japan), Md. Monirul Islam (Bangladesh), Zheny-Yu Wang (MSc student) Summary

Many real-world problems are too large and too complex for a single neural network (NN) to solve alone. A Neural Network (NN) ensemble consisting of several individual NN's has been shown to be able to improve their generalization performance [1].

There has been much work in training NN ensembles [3], in mixtures of experts, and in various boosting and bagging methods. However, all these methods are used to adapt weights in an ensemble. The structure of the ensemble, e.g., the number of NNs in the ensemble, and the structure of individual NNs, e.g., the number of hidden nodes, are all designed manually and fixed during the training process except for one case [2]. While manual design of NNs and ensembles might be appropriate for problems where rich prior knowledge and an experienced NN expert exist, it often involves a tedious trial-and-error process for many real-world problems because rich prior knowledge and experienced human experts are hard to get in practice.

This project studies constructive ensemble learning algorithms [2], where the ensemble structure, NN structure and NN weights are trained incrementally. It will identify the strength and weakness of CNNE [4] and study different approaches to learn NN and ensemble structures automatically. Newly developed algorithms will be tested on both benchmark problems and real world problems from Thales. External Funding

Thales Research and Technology (UK), EPSRC. Selected References

[1] Liu. Y, and Yao. X, "Simultaneous training of negatively correlated neural networks in an ensemble," IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 29(6): 716-725, December 1999.
[2] Islam. Md. Monirul, Yao. X, and Murase. K, "A constructive algorithm for training cooperative neural network ensembles," IEEE Transactions on Neural Networks, 14(4): 820-834, July 2003.
[3] Wang. Z. Y, Yao. X, and Xu. Y, "An Improved Constructive Neural Network Ensemble Approach to Medical Diagnoses," Proc. of the 5th International Conference on Intelligent Data Engineering and Automated Learning (IDEAL'04), Lecture Notes in Computer Science, Vol. 3177, Springer, August 2004, pp.572-577.

Neural Network Ensembles and Co-Evolution for Prediction and Recognition Researchers

Xin Yao, Vineet Khare (PhD student), Yong Liu (Japan), Md. Monirul Islam (Bangladesh), Bernhard Sendhoff (Honda) Summary

Neural Network (NN) ensembles have been shown to perform well for difficult problems [1]. They can be regarded as an approach to automatic divide-and-conquer. This project focuses on the credit assignment issue in co-evolving a hierarchical NN system [3], in which populations of neurons, NNs and NN ensembles learn iteratively to solve a complex problem. Such a co-evolutionary system encompassed both the competitive and cooperative co-evolution. Although this project is done in the context of co-evolving NN ensembles, it has a very wide implication because the credit assignment issue occurs frequently in machine learning. The current work [2] concentrates on different methods to evaluate the fitness of single neurons when the only available information is the training performance of entire NN's. External Funding

Honda Research Institute (Europe), Germany. Selected References

[1] Liu. Y, Yao. X, and Higuchi. T, "Evolutionary Ensembles with Negative Correlation Learning," IEEE Transactions on Evolutionary Computation, 4(4): 380-387, November 2000.
[2] Khare. V, Yao. X, and Sendhoff. B, "Credit Assignment among Neurons in Co-evolving Populations," Accepted by Proc. of the 8th International Conference on Parallel Problem Solving from Nature (PPSN VIII)}, Birmingham, UK, 18-22 September 2004.

Co-evolution of Iterated Prisoner's Dilemma Game Strategies Researchers

Xin Yao, Siang Yew Chong (PhD student), Paul Darwen (Australia), Sung-Bae Cho (South Korea) Summary

The most prominent characteristic of co-evolution is that the fitness of an individual depends on the fitness of other individuals, which can be from the same population or a different population. Co-evolution can be used in both learning and optimisation, although this project is focused on learning iterated prisoner's dilemma (IPD) game strategies. The key research issues on co-evolution include:

Generalisation in co-evolutionary learning,
How to encourage "arms race" in co-evolution,
Interactions/combinations between co-evolution and fitness sharing for automatic modularisation, and
How fitness landscapes (for different individuals) change and interact dynamically in co-evolution.

The key research issues on IPD games include multiple levels of cooperation [1], reputation [2] and noise [3]. External Funding

EPSRC, Overseas Research Students Award from Universities UK. Selected References

[1] Darwen. P, and Yao. X, "Co-Evolution in Iterated Prisoner's Dilemma with Intermediate Levels of Cooperation: Application to Missile Defense," International Journal of Computational Intelligence and Applications, 2(1): 83-107, 2002.
[2] Yao. X, and Darwen. P, "How Important Is Your Reputation in a Multi-Agent Environment," Proc. of the 1999 IEEE Conference on Systems, Man, and Cybernetics, IEEE Press, Piscataway, NJ, USA, pp. II-575 - II-580, October 1999.
[3] Chong. S. Y, and Yao. X, "The Impact of Noise on Iterated Prisoner's Dilemma with Multiple Levels of Cooperation," Proc. of the 2004 Congress on Evolutionary Computation (CEC'04)}, pp.348-355, IEEE Press, Piscataway, NJ, USA, 19-23 June 2004, Portland, Oregon, USA.

Global Optimisation and its Applications in Materials Modelling Researchers

Xin Yao, Chang-Yong Lee (South Korea), Jianguo Lin (Mechanical Engineering), Yong Liu (Japan), Ko-Hsing Liang (China) Summary

Optimisation is ever-present. This project is aimed at functions that are non-differentiable. We have proposed a new evolutionary algorithm, i.e., fast evolutionary programming, which uses Cauchy mutation and its mixture with Gaussian mutation [1]. The algorithm have been improved recently using Levy mutations [2] and applied to solve difficult real-world problems in materials modelling [3]. Self-adaptive schemes have also been studied for adjusting lower bounds for the strategy parameters in evolutionary programming [4]. The focus of this project in the near future will be various real world applications. External Funding

Royal Society (UK), KOSEF (South Korea) Selected References

[1] Yao. X, Liu. Y, and Lin. G, "Evolutionary programming made faster," IEEE Transactions on Evolutionary Computation, 3(2): 82-102, July 1999.
[2] Lee. C. Y, and Yao. X, "Evolutionary programming using the mutations based on the Levy probability distribution," IEEE Transactions on Evolutionary Computation, 8(1): 1-13, January 2004.
[3] Li. B, Lin. J, and Yao. X, "A Novel Evolutionary Algorithm for Determining Unified Creep Damage Constitutive Equations," International Journal of Mechanical Sciences, 44(5): 987-1002, 2002.
[4] Liang. K. H, Yao. X, and Newton. C. S, "Adapting self-adaptive parameters in evolutionary algorithms," Applied Intelligence, 15(3): 171-180, November/December 2001.

Computational Time Complexity of Evolutionary Algorithms Researchers

Xin Yao, Jun He Summary

Evolutionary computation researchers appear to use a rather different approach from "mainstream" computer scientists to analysing evolutionary algorithms. While most computer scientists use computational time or space complexity to characterise an algorithm's performance on a problem, most results about evolutionary algorithms are on algorithm's convergence and convergence rate. Few computational complexity results are available in evolutionary computation in spite of numerous papers on the application of evolutionary algorithms to various combinatorial optimisation problems. It is unclear whether or not evolutionary algorithms are more powerful than existing algorithms on a particular problem in terms of computational complexity. The first aim of this project is to analyse the time complexity of EA's for selected combinatorial optimisation problems. The role of population will be studied in depth. In particular, we will compare (1+1) EA's against (N+N) EA's. The second aim of the project is to obtain insights in order to design efficient EA's for combinatorial optimisation problems. External Funding

EPSRC Selected References

[1] He. J, and Yao. X, "A study of drift analysis for estimating computation time of evolutionary algorithms," Natural Computing, 3(1): 21-35, January 2004.
[2] He. J, and Yao. X, "Towards an analytic framework for analysing the computation time of evolutionary algorithms," Artificial Intelligence, 145(1-2): 59-97, April 2003.
[3] He. J, and Yao. X, "From an Individual to a Population: An Analysis of the First Hitting Time of Population-Based Evolutionary Algorithms," IEEE Transactions on Evolutionary Computation, 6(5): 495-511, October 2002.

Prediction and Analysis of Epileptic Seizures using EEG Data Researchers

Xiaoli Li, Xin Yao, John Jefferys (Medical School, University of Birmingham) Summary

The Electroencephalogram (EEG) has been a very important clinical tool for the evaluation and treatment of epilepsy. The reliable detection of the pre-seizure state can improve anti-epileptic drug management, and could provide early warning to activate an electrical interruption system. This study uses computational intelligence techniques and dynamical system theory to improve the existing EEG monitoring systems, including better prediction accuracy.

In this study we have applied wavelet transform and nonlinear system theory to explore the different indicators of pre-seizure from EEG data. We have also used simple and reliable synchronization measures to describe correlation among different channels for discovering the mechanism of epileptic seizures; and used dynamical similarity measures from complexity theory for improving the prediction accuracy of the seizures.

We have found that high frequency (>100Hz) ripples are associated with the seizure discharge, and found ripples were more common in the run up to seizures. We also found that:

Phase synchronization measures between different areas increased prior to epileptic seizure, and phase synchronization measures at the different frequency bands are independent.
Bispectral and coherence with wavelet transforms discovered the non-linear interaction between two EEG segments; we found fast ripples and gamma waves have a strong interaction during seizures.

External Funding

Wellcome Trust Industrial Projects

Cercia routinely undertakes projects with industrial partners, which tackle issues that our partners face in their day-to-day business. Our endeavour is to provide innovative solutions using CI techniques to solve problems that traditional techniques cannot.

List of Industrial Projects

Force Based Visualisation for Analysis of Student Psychometric Test Results, (Andy Pryke, Colin Frayn , Company Assisted: W3 Insights) (Abstract, hyper link)
A Meta Search Engine for Company Profile Matching, (Colin Frayn, Andy Pryke, Company Assisted: SoundsLike 2D
Salting Route Optimisation with Evolutionary Algorithms (Jin Li, Yong Xu, Company assisted: Entice Technology Ltd)
Analytics Customer Relationship Management with Computational Intelligence (Jin Li, Colin Frayn, Andy Pryke) (Company assisted: Bluespheres)
Computational Intelligence Solutions for Precision Machining and Engineering Sector (Thorsten Schnier, Colin Frayn, Jin Li, Company Assisted: E-9 Ltd)
Computational Intelligence for Sculpture and Creative Arts (Thorsten Schnier, Bob Hendley, Company Assisted: Renn & Thacker Partnership)
Energy Prediction and Outlier Detection (Thorsten Schnier, Company Assisted: E-9 Ltd)
Learning based Process Control In Manufacturing (Xiaoli Li, Xin Yao, Company Assisted: Automation Interfaces Ltd)
Health Monitoring System With Remote Measurement For Electricity, Water Or Gas System (Xiaoli Li, Thorsten Schnier, Company Assisted: Digiterm Ltd, abstract, hyper link)
Application Of Speech Recognition To Rehabilitation Robotics (Xiaoli Li, Company Assisted: Rehabilitation Robotics)
Intelligent Electronic Measuring Instruments For The Automobile Industry (Xiaoli Li, Company Assisted: Solutions Team Ltd)
A Review of CI Techniques with Emphasis on Intelligent Search Algorithms Colin Frayn, Company Assisted: Prism Visual Analysis)
High-Tech Security Solutions for Airports through Smart Tagging (Colin Frayn, John Bullinaria, Peter Ti�o, Company Assisted: UBS Solutions Ltd.)
Prediction of the Properties of Carbohydrate and Protein Structures
Nature-Inspired Algorithms for Visual Arts, (Thorsten Schnier, Company Assisted: Myfanwy Johns ) (Abstract hype link)
A Review of CI Techniques for Financial Application (Colin Frayn, Company assisted: F1 Computing) (Abstract hype link)
Focus Groups
Art and Design Cluster

Force Based Visualisation for Analysis of Student Psychometric Test Results Researchers

Andy Pryke, Colin Frayn Company Assisted

W3 Insights Abstract

W3 Insights is a West Midlands based SME that produces psychometric tests for school children. These tests aim to identify problems regarding attitudes to school and learning, and are used by schools to target interventions.

This project aims to give W3 Insights more powerful analysis of data gathered using their PASS system. Additionally, it will boost the visual appeal of their offerings to clients, and allow more exploratory analysis. The intended result is to increase the value that the W3 Insights consultancy service gives to clients.

Data from the PASS system consists of a number of percentile scores for each pupil. A typical school may apply PASS to several hundred pupils, and a local authority may have thousands of pupils. Cercia have applied their 3D force based data visualisation software to sample data. The software provides an abstract 3D view of multi-dimensional data, in a similar manner to Projection Pursuit, Principal Components Analysis, and Multi Dimensional Scaling. Each pupil is represented as a sphere in a 3D space. Our algorithm clusters together similar pupils, and separates dissimilar ones

In addition to data visualisation, our system uses evolutionary computation to find understandable rules that characterise particular groups. For example, one such rule might be:

"Pupils with Attitude to Teachers in the top 70% have Preparedness in the top 50%" (Applies to 30 pupils, 97% correct)".

The genetic algorithm (GA) would typically consider thousands of potential rules in its search for the most accurate and interesting.

W3 Insights are "very excited" by our approach, and generated a large number of ideas for applications of the software to their data. The company is currently obtaining funding for a project to license and customise the software for their use. References

[1] Pryke, A (1999). Data Mining using Genetic Algorithms and Interactive Visualisation (PhD Thesis), The University of Birmingham.
[2] Pryke, A and Frayn, C. Business Assist Report 2004/5-I-15

A Meta Search Engine for Company Profile Matching Researchers

Colin Frayn, Andy Pryke Company Assisted

SoundsLike 2D Abstract

West Midlands-based entrepreneur Peter Timbrell came to Cercia with this project proposition. The aim is to develop a 'meta search engine' for matching company profiles against one or many company databases, with the aim of identifying potential collaborative partners. The database format could include a number of sources including those in foreign languages. The applications for this project lie in company matching for potential collaboration, takeover or competition analysis. [1]

These studies focussed on a selection of promising technologies that we believe to be relevant to the problems outlined in this project. Firstly, we analysed the market situation for such a company-matching tool. The market for such a package exists, and is worth in the region of a hundred million GBP annually. Currently, the state-of-the-art is poorly defined. Large commercial databases exist and searches can be performed but this is inflexible. Little data is numeric; most of the leading databases contained significant errors and were incomplete. [2] The main problem at the moment is that the matching is very rigid - rather like a bounded, conjunctive rule search instead of a more sensible displacement-based search, which would generate a more valuable (fuzzy) selection. In addition, the standard industry codes (SIC) used to denote industry sectors are more than ten years out of date, therefore missing the entire Internet revolution completely. [3]

We concluded that a full textual analysis solution was not feasible, though several techniques would prove relevant. Most importantly, Bayesian analysis of word frequencies, and latent semantic indexing should be investigated as a means of matching text-based records. For numerical records, some specialised fuzzy matching technique should be employed, allowing a company to relax some of its constraints dynamically in order to accommodate highly accurate matches in other criteria.

We also investigated adapting the Force-Based Clustering and Visualisation techniques developed by Cercia to this purpose. This is software unique to Cercia and, to the best of our knowledge, has not been used before for this purpose. [4] Selected References

[1] Frayn, C, 2004, Cercia Business Assist Report 2003/4-I-4
[2] Cornbill, J., EpiCentre, Coventry. Personal communication.
[3] Johnson, A., Staffordshire Business Link, Personal communication
[4] Ibid. Section

Salting Route Optimisation with Evolutionary Algorithms Researchers

Jin Li, Yong Xu Company Assisted

Entice Technology Ltd Abstract

The optimisation of salting routes is extremely complex and involves numerous interdependent variables or constraints such as the road network, treatment requirements, materials characteristics, vehicle and deport location data, availability of staff and limitations of treatment times.

Traditionally, salting route optimisation is aimed at maximising efficiency in such a way that the distance travelled is minimised. Previously, this has been a manual task, heavily reliant on local knowledge. In practice, highway authorities or local councils have sought either to optimise salting routes for a decade at a time, or to totally avoid the issue. This is partly due to the huge cost of the task. Therefore, it is highly desirable to have a software system available to automatically recommend optimised route solutions in terms of constraints input by users.

Attempts to solve the problem have been made using computational intelligence [1] with success to some extent. Evolutionary computation has been shown to out-perform traditional operations research techniques for similar routing problems [2]. The research of use of both GIS and evolutionary algorithms will provide a superior and robust avenue to route optimisation. It is worth noting that a distinguishing feature of this system is that a weather-related input is treated as an important variable. This could lead to results with a series of ranked thermal routes that vary depending on atmospheric stability. This allows users to make a decision on which route to treat with respect to the daily weather forecast. Selected References

[1] Li, L.Y.O. & Eglese, R.W. (1996). An interactive algorithm for vehicle routing for winter gritting. Journal of the Operational Research Society 47:217-228.
[2] Yao, X. (2003). The evolution of evolutionary computation. Knowledge-based intelligent information and engineering systems Part 1. Proceedings Lecture Notes in Artificial Intelligence 2773:19-20.
[1] Li, J. 2004, Cercia Business Assist Report 2004/5-I-17

Analytics Customer Relationship Management with Computational Intelligence Researchers

Jin Li, Colin Frayn, Andy Pryke Company Assisted

Bluespheres Abstract :

Customer relationship management (CRM) provides customer-oriented services for planning, developing, maintaining, and expanding customer relationships. CRM enables a company to capture a consolidated customer view through multi-channel interactions in a data warehouse. Traditionally, CRM automates processes in sales, marketing, and service functions, aimed at increasing the efficiency of these processes to improve customer satisfaction. With tremendous amounts of data available, the holistic understanding of customer behaviour over time has become possible and highly desirable. However, the existing CRM system could not identify, analyse and predict changes in customer behaviour. Employing computational intelligent techniques such as decision tree algorithms, na�ve Bayesian genetic algorithms, k-means nearest neighbour etc can only spot those behaviour changes. With customer behaviour identified, a company is able to widen customer relationships by:

Revealing descriptive profiles of the best and worst customers
Distinguishing short-term customer from long loyal customers
Optimising customer service with more details in mind
Retaining its customers, minimizing customer churning

Such an analytical CRM system is highly sought-after in markets, especially in manufacturing companies and insurance companies with a large customer base. It will provide a significant and rapid return on investment. Selected references

[1] Li, J. 2004, Cercia Business Assist Report 2003/4-III-9

Computational Intelligence Solutions for Precision Machining and Engineering Sector Researchers

Thorsten Schnier, Colin Frayn, Jin Li Company Assisted

E-9 Ltd Summary

Cercia worked with a local consulting company for this project. One of the clients of the company is a machine and engineering company who is part of the supply chain of Rolls Royce. They had identified a number of issues in the production process, and asked Cercia if we could offer solutions to these problems. Cercia has produced a detailed report, with CI solutions for: Job shop scheduling

Scheduling is a difficult problem, for both evolutionary and conventional algorithms, because generally there are a very large number of possible schedules. CI techniques can be used for automatic scheduling, automatic repair of schedules, and analysis and visualisation of the production process. Factory Floor Plan optimization

From a technological viewpoint, this is a fairly simple problem to solve. The main requirement is the availability of sufficient information from the ERP system. To find a good solution, we would need exact information of all the travel the parts did between machines in a representative period of time (e.g. one year). Batch Size Optimization

The solution for this problem depends on a fairly large amount of additional information. To start with, some cost model is required - for both the possible savings, and the costs of keeping the parts in stock. The other requirement is a model of the client's behaviour - how long is it most likely to take until more items of this part are ordered. Packaging Optimization

This interesting problem could be a bit more technically challenging, depending on the level of detail required. In the simplest scenario, the algorithm would simply optimize a set of rectangular boxes, with the condition that the boxes are large enough to fit the parts it is intended for. The more difficult problem would look into non-rectangular containers, which are designed to support pieces without additional packaging. Here, it would be necessary to compute points of contact between pieces and containers, and the forces generated. Selected References

[1] Schnier. T, Computational Intelligence Solutions for Precision Machining and Engineering Sector Cercia Business Assist Report 2003/04-II-8, 2004
[2] Sarker. R, and Yao. X, "Simuated Annealing for Solving a Manufacturing Batch-Sizing Problem", Journal of Operations and Quantitative Management, 9(1):65-80, March 2003

Computational Intelligence for Sculpture and Creative Arts Researchers

Thorsten Schnier, Bob Hendley Company Assisted

Renn & Thacker Partnership Abstract

Two artists interested in the connection between computer science and art contacted Cercia. Specifically, they were researching possibilities for a sculpture competition, with the potential subject of Charles Darwin.

Cercia was able to demonstrate a number of projects and provide a report, centring on artificial evolution and evolutionary art, with a number of suggestions for use in dynamic and interactive sculpture. We also provided a letter of support for the competition entry, which the company later won. The company has also joined Cercia's bids for network funding, and will be part of Cercia's design network.

There are a number of potential CI techniques that can be used to create innovative sculptures. In general, these techniques can be applied in two different ways. Firstly, it is possible to use a number of CI techniques to generate sculptures and other artwork. In this case, the CI techniques are used by the artist/designer to help them open up and explore new design spaces. The second suggested use is as part of the sculpture or artwork. Here, the CI techniques are built into the sculpture, as part of a dynamic and/or interactive artwork. For both forms of use, a number of different CI techniques are available that Cercia has expertise in. Selected References

[1] Schnier. T, Computational Intelligence for Sculpture and Creative Arts, Cercia Business Assist Report 2003/04-II-7, 2004
[2] Schnier. T, Evolved Representations and Their Use in Computational Creativity, Ph.D. Thesis, University of Sydney, 1999

Energy Prediction and Outlier Detection Researchers

Thorsten Schnier Company Assisted

E-9 Ltd Abstract

Cercia's client here, was a small consulting company. As part of its product range, this company offers a service to monitor and analyze a client's electricity consumption, and display the results in a web-based interface. They wanted to improve the analytical aspect of its product, and wanted to know if we could analyze the data collected, and automatically extract the contribution of individual 'units' (e.g. particular machines, A/C, office use) from the consumption data.

The client supplied Cercia with a number of sample data streams, taken at a number of commercial sites at 30-minute intervals. We performed extensive analysis of the data, tested a number of possible algorithms on the data, to see how much 'explanation' of the data can be derived using a purely data-driven approach.

We first established that k-means clustering is able to cluster the data for each particular day into a different group depending on the work pattern of that day. To further analyze the data, we used an evolutionary algorithm to derive an optimal set of time-based partitions that allows us to describe any day within a particular cluster as a series of 'phases'. The start and end times of the phases is the same for all days within one cluster (e.g. low consumption over night, followed by a warm-up period for the morning shift from 7.30am to 10.30am, a period of intensive energy usage until 12am, a short drop for shift change until 12.30 am and so on), but the consumption values (taken as average over each phase) will be different for any day. The algorithm essentially performs a data reduction, reducing the 48 samples each day into a small number of values (e.g. 6) that can meaningfully be compared within each day.

The work established that although data alone is insufficient to attribute particular consumption to any particular process within the site, it is still possible to derive some meaningful values if the days are first clustered into groups with similar patterns. Selected References

[1] Schnier. T, Energy Prediction and Outlier Detection, Cercia Business Assist Report 2002/03-IV-2, 2004

Learning based Process Control In Manufacturing Researchers

Xiaoli Li, Xin Yao Company Assisted

Automation Interfaces Ltd Abstract

The heat treatment process in manufacturing needs a very fast, auto-tuning process controller. Sensor-based control algorithms cannot meet this need. We proposed fuzzy genetic algorithms (FGA) and iterative learning methods to construct a fast process controller. A fuzzy genetic algorithm can automatically construct a model of the controlled system based on sensed data. Iterative learning can rapidly adjust the input of process controller based on the feedback error to improve control accuracy. Health Monitoring System With Remote Measurement For Electricity, Water Or Gas System Researchers

Xiaoli Li, Thorsten Schnier Company Assisted

Digiterm Ltd Abstract

The goal of a monitoring system is to discover the fault from the collected data, then to provide fault prediction in a simple format, and to recommend preventative maintenance. We proposed a new monitoring system based on an evolutionary computation - fuzzy neural network and wavelet transform, which can extract some features from the collected data, and discover some transient phenomena in the power, water and gas systems. Application Of Speech Recognition To Rehabilitation Robotics Researchers

Xiaoli Li Company Assisted

Rehabilitation Robotics Summary

We investigated a fuzzy speech recognition method to identify the requirements from the speaker, then generated control signal commands for a patient rehabilitation application. Intelligent Electronic Measuring Instruments For The Automobile Industry Researchers

Xiaoli Li Company Assisted

Solutions Team Ltd Abstract

We investigated self-calibration and self-diagnosis techniques to enhance operation of new electronic measuring instruments for use in automotive applications. We can use an imperfectly calibrated measuring instrument and one or more imperfectly calibrated measurement artifact to improve calibration of the instrument and the artifact(s). Using electronic measuring instruments to collect data from automobile, an intelligent measurement system should know whether the output is a reasonable value or not. Self-diagnosis technique can give a solution on this problem. A Review of CI Techniques with Emphasis on Intelligent Search Algorithms Researchers

Colin Frayn Company Assisted

Prism Visual Analysis Abstract

We explored the feasibility of designing intelligent travel-search software. Specifically, two aspects were investigated. First, identifying travel patterns and suggesting destinations that people might not have considered based on the type of holiday that people with similar tastes had taken before. Second, enabling general destination specification - searching a database of destinations or holiday types based on keywords or a general textual specification rather than a fixed, rigid set of criteria. High-Tech Security Solutions for Airports through Smart Tagging Researchers

Colin Frayn, John Bullinaria, Peter Ti�o Company Assisted

UBS Solutions Abstract

We provided UBS Solutions with a feasibility study for developing a system to identify unusual movement patterns of people in an airport. The proposed system had a database that stored historical transition information. The aim of the system was to learn ordinary, safe patterns of movement within the airport so that the current actions of the passengers in the airport could be analysed and any unusual movements highlighted. Prediction of the Properties of Carbohydrate and Protein Structures Researchers

Yong Xu, Xiaoli Li Company Assisted

Chembiotech Limited Abstract

Chembiotech is an R&D lab specialising in the design of new carbohydrates and proteins. Cercia has provided Chembiotech with an overview of CI methods that can be used to predict and model chemical and physical properties of compounds using chemical structure information. Nature-Inspired Algorithms for Visual Arts Researchers

Thorsten Schnier Company Assisted

Myfanwy Johns Abstract

Cercia has worked closely with Myfanwy Johns, a graphical design artist, to explore how CI techniques can be used to generate new, interesting pieces of graphical art. A Review of CI Techniques for Financial Application Researchers

Colin Frayn Company Assisted

F1 Computing

We have provided F1 computing, a software design house, an exhaustive overview of the applications of CI techniques to assess credit card applications in financial institutions. Focus Groups

As part of Cercia's aim to be an international leader in applied research and knowledge transfer of computational intelligence techniques, we are in the process of forming groups that will help focus our research expertise in three broad areas:

Adaptive Optimisation
Machine Learning and Data mining
Creative Design

Cercia has already begun setting up one such group, and expects to develop more in the coming year. Art and Design Cluster

We are developing a local cluster of scientists, artists, designers, and industrialists that will focus on the interaction between Computational Intelligence and Art and Design. This project will focus on the creative process in arts, music, design and culture studies and investigate the two-way interaction between science and engineering and the creative arts. We will study how advanced computer science and engineering methods and techniques can be used to encourage and facilitate the creative process. We will also explore how principles and ideas from creative arts and design can be used to inspire better and more human-friendly software and hardware systems. The cluster will have three main tasks:

To bring people together to identify possibilities for collaboration;
To encourage and support the collaborations;
And to explore new research directions and topics.

The cluster activities consist of workshops, seminars and exhibitions as well as an online forum. External Funding

Network grant worth �50,000 from AHRB and EPSRC as part of the 'Designing for the 21st Century' program.