Research Interests:
ABSTRACT In real-world computational engineering design, it is often necessary to consider a large number of design parameters to describe and model the product under investigation, and a large number of objective functions to express the... more
ABSTRACT In real-world computational engineering design, it is often necessary to consider a large number of design parameters to describe and model the product under investigation, and a large number of objective functions to express the strongly competing performance metrics. As a result, when we successfully perform multi-objective optimization studies, we produce a large number of data points in a high-dimensional space. In order to develop understanding of the complexities in such problems, we need to simultaneously visualise the design parameters and objective functions spaces and reveal the patterns of optimum performance, but also the discontinuous areas of infeasibility. By achieving this, we can potentially benefit in many ways. Firstly, the information that is extracted from such analysis can explicitly support human-in-the-loop computational design methodologies. Secondly, multi-fidelity optimization approaches can be guided and further enhanced with trust region approaches. Thirdly, identify the causes and reasons for the activation of the physical mechanisms that are responsible for improved, but also degraded, flow behavior in compressors design. And finally, the identification of design rules and characterization of the complexity in real-world engineering design. We use the Parallel Coordinates technique to manage the visualization and interaction of such hyper-planes. The illustrative case study used is the 3D aerodynamic design of compressor blades, considering 26 design parameters for the geometry management and 2 objective functions to model two of the critical flow characteristics (c1 - entropy generation, c2 - blockage).
Problems such as resource and service discovery models, load balancing and scheduling, brokering are eminent in grid systems due to bottlenecks such as bandwidth and network traffic in the underlying communication infra- structures and... more
Problems such as resource and service discovery models, load balancing and scheduling, brokering are eminent in grid systems due to bottlenecks such as bandwidth and network traffic in the underlying communication infra- structures and their associated costs in fabricating a scal- able and cost effect grid services infrastructure. The pri- mary goals of this paper is to apply grid based coalition formation concepts to agents: add efficient load balancing and scheduling (A 3p viLoad Scheduler) schemes; provide a replacement solution to resource discovery models by applying application oriented directory services; minimize message passing of state information updates and eco- nomic brokering services to the agent aware adhoc p2p virtual interconnect grid computing system or A 3p vigrid system.
Research Interests:
Due to the gradient-free mechanism, flexibility, high local optima avoidance, and simplicity, meta-heuristics have been reliable alternatives to conventional optimisation techniques over the course of last two decades. This has resulted... more
Due to the gradient-free mechanism, flexibility, high local optima avoidance, and simplicity, meta-heuristics have been reliable alternatives to conventional optimisation techniques over the course of last two decades. This has resulted in the application of such techniques in diverse branches of science and technology. Despite all the successful applications, meta-heuristics are less effective in real-time applications where there is a need to find the optimal solutions instantly due to the need for a large number of function evaluations. This paper investigates the effectiveness of meta-heuristics in modelling hands for recognising hand gestures. Several well-known and recent algorithms have been utilised to find an optimal shape for a 3D model of the hand. Qualitative and quantitative results have been collected to see how well meta-heuristics perform in this field. Firstly, the results show that a free model of the hand can be very expensive to optimise: a constrained model is essential to reduce the search space. Secondly, the results show that population-based algorithms are more suitable rather than individual-based mainly because of the presence of a large number of local solutions. Thirdly, despite the accuracy of the optimal model obtained using population-based algorithms, the run time is an issue which should be considered. Finally, several recommendations are made for reducing the run time of meta-heuristics and making them more practical in the field of gesture detection.
Research Interests:
Artificial Neural Networks (ANNs) have often been used to model objective functions for Multi-Objective Particle Swarm Optimisation (MOPSO); alternatively MOPSO has been used to aid in training ANNs. In previous work we instead used an... more
Artificial Neural Networks (ANNs) have often been used to model objective functions for Multi-Objective Particle Swarm Optimisation (MOPSO); alternatively MOPSO has been used to aid in training ANNs. In previous work we instead used an ANN to guide optimisation by deciding if a trial solution was worthy of full evaluation. In this work we introduce Active Learning to the ANN-guided MOPSO. This is done by using a dynamic subset of particles from the MOPSO swarm to classify locations that are likely to be on the boundary between feasible and infeasible space. As a case study we sought to optimise the shape of an airfoil to minimise drag and maximise lift.We investigated the effect of allowing up to 20 particles from the swarm to be used for Active Learning. Our analysis showed the addition of Active Learning resulted in an increase in performance where an initial archive for training was available. However if an initial archive was not available then Active Learning performed at best equal to non-Active Learning and often worse, in some cases showing poorer performance than an unguided MOPSO.
Research Interests: Computer Science and IEEE
Research Interests:
Research Interests:
Research Interests:
This article proposes a method for designing electromagnetic compatibility shielding enclosures using a peer-to-peer based distributed optimization system based on a modified particle swarm optimization algorithm. This optimization system... more
This article proposes a method for designing electromagnetic compatibility shielding enclosures using a peer-to-peer based distributed optimization system based on a modified particle swarm optimization algorithm. This optimization system is used to obtain optimal solutions to a shielding enclosure design problem efficiently with respect to both electromagnetic shielding efficiency and thermal performance. During the optimization procedure it becomes evident that
Research Interests: Engineering, Distributed Computing, Electromagnetic Compatibility, Computer Aided Design, Engineering Optimization, and 9 moreMathematical Sciences, Computer Model, Thermal Performance, Peer to Peer, Design optimization, Large Scale, Distributed Optimization, Particle Swarm Optimizer, and Optimal Solution
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
ABSTRACT
This paper describes a novel tool called Nimrod/O that allows a user to run an arbitrary
Research Interests:
Research Interests:
ABSTRACT The run time for many optimisation algorithms, particularly those that explicitly consider mul- tiple objectives, can be impractically large when applied to real world problems. This paper reports an investigation into the... more
ABSTRACT The run time for many optimisation algorithms, particularly those that explicitly consider mul- tiple objectives, can be impractically large when applied to real world problems. This paper reports an investigation into the behaviour of Multi-Objective Particle Swarm Optimisation (MOPSO), which seeks to reduce the number of objective function evaluations needed, with- out degrading solution quality. By restricting archive size and strategically reducing the trial solution population size, it has been found the number of function evaluations can be reduced by 66.7% without significant reduction in solution quality. In fact, careful manipulation of algorithm operating parameters can even significantly improve solution quality.
Research Interests:
Fundamental antenna limits of the gain-bandwidth product are derived from polarizability calculations. This electrostatic technique has significant value in many antenna evaluations. Polarizability is not available in closed form for most... more
Fundamental antenna limits of the gain-bandwidth product are derived from polarizability calculations. This electrostatic technique has significant value in many antenna evaluations. Polarizability is not available in closed form for most antenna shapes and no commercial electromagnetic packages have this facility. Numerical computation of the polarizability for arbitrary conducting bodies was undertaken using an unstructured triangular mesh over the surface of 2D and 3D objects. Numerical results compare favourably with analytical solutions and can be implemented efficiently for large structures of arbitrary shape.
Research Interests:
Research Interests:
ABSTRACT The issue of studying the effect of fixing the length of the selected feature subsets using ant colony optimization (ACO) has not yet been studied. This paper addresses this concern by demonstrating four points that are: 1)... more
ABSTRACT The issue of studying the effect of fixing the length of the selected feature subsets using ant colony optimization (ACO) has not yet been studied. This paper addresses this concern by demonstrating four points that are: 1) determining the optimal feature subset, 2) determining the length of the subsets in ACO for subset selection problems, 3) different stopping criteria when solving feature selection by ACO, and 4) experiments on an ACO algorithm for feature selection problems using artificial and real-world datasets in two cases fixing and not fixing the length of the selected feature subsets with the use of a support vector machine (SVM) classifier. The results showed that not fixing the length of the selected feature subsets is better than fixing the length of the selected feature subsets in terms of the classifier accuracy in seven datasets out of ten.