Differential Evolution (DE) is a highly successful population based global optimisation algorithm, commonly used for solving numerical optimisation problems. However, as the complexity of the objective function increases, the wall-clock run-time of the algorithm suffers as many fitness function evaluations must take place to effectively explore the search space. Due to the inherently parallel nature of the DE algorithm, graphics processing units (GPU) have been used to effectively accelerate both the fitness evaluation and DE algorithm. This work reviews the main architectural choices made in the literature for GPU based DE algorithms and introduces a new GPU based numerical optimisation benchmark to evaluate and compare GPU based DE algorithms.

This paper presents a visualisation tool (ECvis) that aids the development of population based numerical optimisation algorithms such as genetic algorithms and differential evolution. The tool provides a simple interface with three modes: A Density mode that allows the user to quickly view the distribution and density of the population throughout the fitness and search space for high dimensional problems. This provides the ability to quickly establish where the population is clustering, which can indicate potential local and global minima; A Statistical mode that allows the user to visualise the individuals that have statistical significance in the population and their location in the search space; A Ranges mode that provides the user with a windowed average of the minimum, maximum and median value of each parameter of the population and whether the range of each parameter has changed since the previous window. This allows the user to see whether the population is exhibiting exploration or exploitation properties, as well as convergence properties such as the variance of each parameter. As examples of the usefulness of ECvis, two well known, high dimensional functions are optimised using differential evolution with ECvis being used to provide information on the performance of the optimisation.

When a deterministic search approach is too costly, such as for non-deterministic polynomial-hard problems, finding near-optimal solutions with approximation algorithms, such as the genetic algorithm, is the only practical approach to reduce the execution time. In this paper, we exploit the capability of graphics processing units (GPU), specifically Nvidia's CUDA platform, to accelerate the genetic algorithm by modifying the evolutionary operations to fit the hardware architecture. This has allowed us to achieve significant computational speedups compared to the non-GPU counterparts.

The maintenance of an existing large road network is a key focus area for road authorities around the world. The pressures associated with the ever-increasing road network and often shrinking budgets means that it is essential that road authorities invest maintenance budgets wisely. In line with this objective, most road authorities’ employee a Pavement Management System (PMS) to assist in making maintenance decisions. PMSs must solve a very large optimization problem involving thousands of road segments with multiple possible treatments. There is a wide range in the cost of these treatments and also in the magnitude and duration of their improvement. The optimization problem is to identify a minimum cost, 20-year maintenance program that ensures all segments are maintained at an acceptable level (which varies depending on factors such as the amount of traffic and the type of traffic). In addition to the 20-year overall budget, there are yearly budgets constraints which must be met and many other constraints such as the availability of staff and machinery. Previous research has shown significant benefit arises from the adoption of a genetic algorithm-based PMS. This paper builds on this research through the application and evaluation of a tailored, parallel genetic algorithm within a PMS. A tailored genetic algorithm is evaluated using a real-world road network of 1,335 road segments executed using 12 processing units with annual budgets ranging between $40 and $50 million. Over a total of 174 trials, the tailored genetic algorithm was 46% more successful than a standard genetic algorithm at producing an optimised program of works that satisfied all budget constraints, typically with a lower overspend.

All road authorities are required to make sound maintenance investment decisions to maximise value from available budgets. As an indication of the complexity of this task, the schedule of pavement maintenance and rehabilitation for a small pavement network consisting of 200 segments, with four treatment alternatives over a planning period of five years has (2004)5 = 1.05 * 1046 possible alternatives. This study investigates the number and quality of solutions obtained by adding additional computing resources to a budget constrained implementation of a Parallel Genetic Algorithm based pavement management treatment scheduling system.

The maximum k -plex problem is an important, computationally complex graph based problem. In this study an effective k -plex local search (KLS) is presented for solving this problem on a wide range of graph types. KLS uses data structures suitable for the graph being analysed and has mechanisms for preventing search cycling and promoting search diversity. State of the art results were obtained on 121 dense graphs and 61 large real-life (sparse) graphs. Comparisons with three recent algorithms on the more difficult graphs show that KLS performed better or as well as in 93% of 332 significant k -plex problem instances investigated achieving either larger average k -plex sizes (including some new results) or, when these were equivalent, lower CPU requirements.

The maximum k-plex problem is an important, computationally complex graph based problem. In this study an effective k-plex local search (KLS) is presented for solving this problem on a wide range of graph types. KLS uses data structures suitable for the graph being analysed and has mechanisms for preventing search cycling and promoting search diversity. State of the art results were obtained on 121 dense graphs and 61 large real-life (sparse) graphs. Comparisons with three recent algorithms on the more difficult graphs show that KLS performed better or as well as in 93% of 332 significant k-plex problem instances investigated achieving either larger average k-plex sizes (including some new results) or, when these were equivalent, lower CPU requirements.

The linear sum assignment problem is a fundamental combinatorial optimisation problem and can be broadly defined as: given an \(n \times m, m \ge n\) benefit matrix \(B = (b_{ij})\), matching each row to a different column so that the sum of entries at the row-column intersections is maximised. This paper describes the application of a new fast heuristic algorithm, Asymmetric Greedy Search, to the asymmetric version (\(n \ne m\)) of the linear sum assignment problem. Extensive computational experiments, using a range of model graphs demonstrate the effectiveness of the algorithm. The heuristic was also incorporated within an algorithm for the non-sequential protein structure matching problem where non-sequential alignment between two proteins, normally of different numbers of amino acids, needs to be maximised.

A subspace code of length \(n\) over the finite field \(\mathbb {F}_q\) is a collection of subspaces of the \(n\)-dimensional vector space \(\mathbb {F}_q^n\). Subspace codes are applied to a number of areas such as noncoherent linear network coding and linear authentication. A challenge in the research of subspace codes is to construct large codes with prescribed code parameters, such that the codes have the maximum number of codewords, or the number of codewords is larger than that of previously known codes. In the literature, a general method was proposed for the construction of large constant-dimension subspace codes based on integer linear programming. In this work, making use of an optimization approach for finding the maximum independent set of a graph, a procedure is developed for constructing large subspace codes. The procedure, in some cases, outperforms the existing approach based on integer linear programming, and finds new subspace codes that have more codewords than existing codes.

Given a graph, the critical node detection problem can be broadly defined as identifying the minimum subset of nodes such that, if these nodes were removed, some metric of graph connectivity is minimised. In this paper, two variants of the critical node detection problem are addressed. Firstly, the basic critical node detection problem where, given the maximum number of nodes that can be removed, the objective is to minimise the total number of connected nodes in the graph. Secondly, the cardinality constrained critical node detection problem where, given the maximum allowed connected graph component size, the objective is to minimise the number of nodes required to be removed to achieve this. Extensive computational experiments, using a range of sparse real-world graphs, and a comparison with previous exact results demonstrate the effectiveness of the proposed algorithms.

This paper presents and compares a tree-based road treatment scheduling system with the more traditional rule-based system using a real-world road network of 1,335 road segments. The tree-based road treatment scheduling system effectively evaluates if a particular treatment should be applied in a later year instead of the current year or if an alternative ‘holding’ treatment should be applied to minimise expenditure. As a consequence, the tree-based road treatment scheduling system evaluates all feasible solutions as compared to the traditional rule-based system, which is basically a greedy search that applies rules that trigger treatments based on intervention levels in a predefined hierarchical order, which produces a single solution and does not evaluate all feasible solutions. The performance of the tree-based road treatment scheduling system was evaluated on road networks of up to 25,000 segments. For a 1,335 road segment network, the work program produced by the tree-based system was costed at 82.3% of that produced by the rule-based system, whilst complying with all stipulated maximum condition constraints.

Motivation: The three dimensional tertiary structure of a protein at near atomic level resolution provides insight alluding to its function and evolution. As protein structure decides its functionality, similarity in structure usually implies similarity in function. As such, structure alignment techniques are often useful in the classifications of protein function. Given the rapidly growing rate of new, experimentally determined structures being made available from repositories such as the Protein Data Bank, fast and accurate computational structure comparison tools are required. This paper presents SPalignNS, a non-sequential protein structure alignment tool using a novel asymmetrical greedy search technique. Results: The performance of SPalignNS was evaluated against existing sequential and non-sequential structure alignment methods by performing trials with commonly used datasets. These benchmark datasets used to gauge alignment accuracy include (i) 9538 pairwise alignments implied by the HOMSTRAD database of homologous proteins; (ii) a subset of 64 difficult alignments from set (i) that have low structure similarity; (iii) 199 pairwise alignments of proteins with similar structure but different topology; and (iv) a subset of 20 pairwise alignments from the RIPC set. SPalignNS is shown to achieve greater alignment accuracy (lower or comparable root-mean squared distance with increased structure overlap coverage) for all datasets, and the highest agreement with reference alignments from the challenging dataset (iv) above, when compared with both sequentially constrained alignments and other non-sequential alignments. Availability and implementation: SPalignNS was implemented in C++. The source code, binary executable, and a web server version is freely available at: http://sparks-lab.org Contact: yaoqi.zhou@griffith.edu.au

Treatment selection techniques used in Pavement Management Systems often rely on predefined priorities outlined by state road agency which are subject to priority and engineering judgement. These techniques can be implemented over multi-period planning horizons however doesn't necessarily provide the best possible works program. This paper presents an overview of Pavement Management Systems, then describes and analyses the four main categories of treatment selection techniques employed within these systems. An overview of the Pavement Management System implemented by the Queensland Department of Transport and Main Roads and in particular it's process of treatment selection is outlined. In addition a comparative analysis is then undertaken with each of these treatment selection categories.

Java Programming Laboratory (JPL) is a cloud based, integrated learning environment used for teaching introductory Java programming at Griffith University, Australia. JPL incorporates a number of features found in other successful programming learning environments and builds upon them with a range of innovative features. Learning to program is a threshold concept area for ICT students and success in this endeavour is paramount for student progression and retention. JPL provides a database that tracks individual students' successes and progression through scaffolded programming exercises and assessment items and gives students immediate feedback on their use of programming language syntax and correctness of problem solutions. A data querying and visualisation facility allows analysis of the database to provide real-time performance indicators from the overall course / problem level down to the individual student / specific problem level. Programming instructors and curriculum designers will find that this facility allows a responsive approach to student engagement, assistance and progression; as well as course problem tuning in a just-in-time manner.

Java Programming Laboratory (JPL) is a cloud based learning environment used for teaching object-oriented programming at Griffith University, Australia. JPL incorporates a number of features found in other successful programming learning environments and builds upon them with a range of innovative features. JPL provides a database that tracks individual students’ successes and progression through scaffolded programming exercises and assessment items and gives students immediate feedback on their use of programming language syntax and correctness of problem solutions. A data querying and visualisation facility allows analysis of the database to provide real-time performance indicators from the overall course / problem level down to the individual student / specific problem level. Programming instructors and curriculum designers will find that this facility allows a responsive approach to student engagement, assistance and progression; as well as course problem tuning in a just-in-time manner.

The advent of desktop multi-core computers has dramatically improved the usability of parallel algorithms which, in the past, have required specialised hardware. This paper introduces cooperating local search (CLS), a parallelised hyper-heuristic for the maximum clique problem. CLS utilises cooperating low level heuristics which alternate between sequences of iterative improvement, during which suitable vertices are added to the current clique, and plateau search, where vertices of the current clique are swapped with vertices not in the current clique. These low level heuristics differ primarily in their vertex selection techniques and their approach to dealing with plateaus. To improve the performance of CLS, guidance information is passed between low level heuristics directing them to particular areas of the search domain. In addition, CLS dynamically reconfigures the allocation of low level heuristics to cores, based on information obtained during a trial, to ensure that the mix of low level heuristics is appropriate for the instance being optimised. CLS has no problem instance dependent parameters, improves the state-of-the-art performance for the maximum clique problem over all the BHOSLIB benchmark instances and attains unprecedented consistency over the state-of-the-art on the DIMACS benchmark instances.

Proteins are one of the most vital macromolecules on the cellular level. In order to understand the function of a protein, its structure needs to be determined. For this purpose, different computational approaches have been introduced. Genetic algorithms can be used to search the vast space of all possible conformations of a protein in order to find its native structure. A framework for design of such algorithms that is both generic, easy to use and performs fast on distributed systems may help further development of genetic algorithm based approaches. We propose such a framework based on a parallel master-slave model which is implemented in C++ and Message Passing Interface. We evaluated its performance on distributed systems with a different number of processors and achieved a linear acceleration in proportion to the number of processing units.

This paper presents the results obtained using an unbiased Population Based Search (PBS) for optimising Morse atomic clusters. PBS is able to repeatedly obtain all putative global minima for Morse clusters in the range 5 ≤ N ≤ 80, N = 147,ρ = 3,6,10, 14, as reported in the Cambridge Cluster Database. In addition, putative global minima have been established for Morse clusters in the range 81 ≤ N ≤ 146,ρ = 14. The PBS algorithm incorporates and extends key techniques that have been developed in other cluster optimisation algorithms over the last decade. Of particular importance are the use of cut and paste operators, structure niching and a new operator, Directed Optimisation, which extends the previous concept of directed mutation. In addition, PBS is able to operate in a parallel mode for optimising larger clusters.

The uncapacitated facility location problem is one of finding the minimum cost subset of m facilities, where each facility has an associated establishment cost, to satisfy the demands of n users where the cost of satisfying each user from all possible facilities is known. In this paper, PBS, a population based metaheuristic for the uncapacitated facility location problem is introduced. PBS uses a genetic algorithm based meta-heuristic, primarily based on cut and paste crossover and directed mutation operators, to generate new starting points for a local search. For larger uncapacitated facility location instances, PBS is able to effectively utilise a number of computer processors. It is shown empirically that PBS achieves state-of-the-art performance for a wide range of uncapacitated facility location benchmark instances.