Dr Kyle Harrison

The article presents two variants of the project portfolio selection and scheduling problem (PPSSP). The primary objective of the PPSSP is to maximise the total portfolio value through the selection and scheduling of a subset of projects subject to various operational constraints. This article describes two recently-proposed, generalised models of the PPSSP [1,2] and proposes a set of synthetically generated problem instances for each. These datasets can be used by researchers to compare the performance of heuristic and meta-heuristic solution strategies. In addition, the Python program used to generate the problem instances is supplied, allowing researchers to generate new problem instances.

Future Force Design (FFD) is a strategic planning activity that decides the programming of defence capability options. This is a complex problem faced by the Australian Department of Defence (DoD) and requires the simultaneous selection and scheduling of projects. Specifically, this is a NP-hard problem known as the Project Portfolio Selection and Scheduling Problem (PPSSP). While the PPSSP is a complex problem itself, its complexity is further increased when coupled with the additional characteristics that arise in the context of defence-oriented planning, such as long planning periods and complex operational constraints. As a result, many previous studies examined only a small number of projects over a short planning period and are largely unsuitable for the scale required in the defence sector. To address this issue, two primary contributions are made in this paper. Firstly, this study describes a complex practical PPSSP, inspired by the FFD process, and develops a corresponding mathematical model. Problem instances are derived from real-world, publicly-available defence data. Secondly, to address instances of the problem, two existing meta-heuristics are considered and a hybrid, multi-population approach is proposed. Results are compared against those attained by a commercial exact solver and indicate that there is no statistically significant difference in performance between the proposed multi-population approach and the exact solver. A key benefit of the proposed meta-heuristic approach is that its run time is not significantly influenced by the complexity of the problem instance. Additionally, many interesting practical insights regarding the solution of selection and scheduling problems are uncovered.

The problem of designing an effective future defense force is quite complex and challenging. One methodology that is often employed in this domain is portfolio optimization, whereby the objective is to select a diverse set of assets that maximize the return on investment. In the defense context, the return on investment is often measured in terms of the capabilities that the investments will provide. While the field of portfolio optimization is well established, applications in the defense sector pose unique challenges not seen in other application domains. However, the literature regarding portfolio optimization for defense applications is rather sparse. To this end, this paper provides a structured review of recent applications and identifies a number of areas that warrant further investigation.

This paper presents an investigation into the short-term versus long-term performance of various particle swarm optimization (PSO) control parameter configurations. While evidence suggests that the best PSO parameter values to employ are time-dependent, this paper provides an in-depth examination of a small set of parameter values to provide a more concrete quantification of the performance degradation observed with specific control parameter configurations over time. Given that the short-term performance is not necessarily indicative of long-term performance, this paper proposes that machine learning techniques be used to build predictive models based on two easily-observable landscape characteristics. Finally, using the predictive models as a basis, this paper also proposes a parameter-free PSO algorithm, which performs on par with other top-performing PSO variants, namely the three best performing static PSO configurations, particle swarm optimization with time-varying acceleration coefficients (PSO-TVAC), and particle swarm optimization with improved random constants (PSO-iRC).

Identifying critical nodes in complex networks has become an important task across a variety of application domains. The Critical Node Detection Problem (CNDP) is an optimization problem that aims to minimize pairwise connectivity in a graph by removing a subset of K nodes. Despite the CNDP being recognized as a bi-objective problem, until now only single-objective problem formulations have been proposed. In this paper, we propose a bi-objective version of the problem that aims to maximize the number of connected components in a graph while simultaneously minimizing the variance of their cardinalities by removing a subset of K nodes. We prove that our bi-objective formulation is distinct from the CNDP, despite their common motivation. Finally, we give a brief comparison of six common multi-objective evolutionary algorithms using sixteen common benchmark problem instances, including for the node-weighted CNDP. We find that of the examined algorithms, NSGAII generally produces the most desirable approximation fronts.

Particle swarm optimization (PSO) is a population-based, stochastic search algorithm inspired by the flocking behaviour of birds. The PSO algorithm has been shown to be rather sensitive to its control parameters, and thus, performance may be greatly improved by employing appropriately tuned parameters. However, parameter tuning is typically a time-intensive empirical process. Furthermore, a priori parameter tuning makes the implicit assumption that the optimal parameters of the PSO algorithm are not time-dependent. To address these issues, self-adaptive particle swarm optimization (SAPSO) algorithms adapt their control parameters throughout execution. While there is a wide variety of such SAPSO algorithms in the literature, their behaviours are not well understood. Specifically, it is unknown whether these SAPSO algorithms will even exhibit convergent behaviour. This paper addresses this lack of understanding by investigating the convergence behaviours of 18 SAPSO algorithms both analytically and empirically. This paper also empirically examines whether the adapted parameters reach a stable point and whether the final parameter values adhere to a well-known convergence criterion. The results depict a grim state for SAPSO algorithms; over half of the SAPSO algorithms exhibit divergent behaviour while many others prematurely converge.

The particle swarm optimization (PSO) algorithm is a stochastic search technique based on the social dynamics of a flock of birds. It has been established that the performance of the PSO algorithm is sensitive to the values assigned to its control parameters. Many studies have examined the long-term behaviours of various PSO parameter configurations, but have failed to provide a quantitative analysis across a variety of benchmark problems. Furthermore, two important questions have remained unanswered. Specifically, the effects of the balance between the values of the acceleration coefficients on the optimal parameter regions, and whether the optimal parameters to employ are time-dependent, warrant further investigation. This study addresses both questions by examining the performance of a global-best PSO using 3036 different parameter configurations on a set of 22 benchmark problems. Results indicate that the balance between the acceleration coefficients does impact the regions of parameter space that lead to optimal performance. Additionally, this study provides concrete evidence that, for the examined problem dimensions, larger acceleration coefficients are preferred as the search progresses, thereby indicating that the optimal parameters are, in fact, time-dependent. Finally, this study provides a general recommendation for the selection of PSO control parameter values.

Complex networks are often characterized by their statistical and topological network properties such as degree distribution, average path length, and clustering coefficient. However, many more characteristics can also be considered such as graph similarity, centrality, or flow properties. These properties have been utilized as feedback for algorithms whose goal is to ascertain plausible network models (also called generators) for a given network. However, a good set of network measures to employ that can be said to sufficiently capture network structure is not yet known. In this paper we provide an investigation into this question through a meta-analysis that quantifies the ability of a subset of measures to appropriately compare model (dis)similarity. The results are utilized as fitness measures for improving a recently proposed genetic programming (GP) framework that is capable of ascertaining a plausible network model from a single network observation. It is shown that the candidate model evaluation criteria of the GP system to automatically infer existing (man-made) network models, in addition to real-world networks, is improved.

Particle swarm optimization (PSO) is a population-based, stochastic optimization technique inspired by the social dynamics of birds. The PSO algorithm is rather sensitive to the control parameters, and thus, there has been a significant amount of research effort devoted to the dynamic adaptation of these parameters. The focus of the adaptive approaches has largely revolved around adapting the inertia weight as it exhibits the clearest relationship with the exploration/exploitation balance of the PSO algorithm. However, despite the significant amount of research efforts, many inertia weight control strategies have not been thoroughly examined analytically nor empirically. Thus, there are a plethora of choices when selecting an inertia weight control strategy, but no study has been comprehensive enough to definitively guide the selection. This paper addresses these issues by first providing an overview of 18 inertia weight control strategies. Secondly, conditions required for the strategies to exhibit convergent behaviour are derived. Finally, the inertia weight control strategies are empirically examined on a suite of 60 benchmark problems. Results of the empirical investigation show that none of the examined strategies, with the exception of a randomly selected inertia weight, even perform on par with a constant inertia weight.

It is well known that appropriately configuring the control parameters for computational intelligence algorithms is a challenging problem. Thus, analysis of the configuration space can provide critical insights towards designing more effective parameter tuning strategies. Recently, the concept of parameter configuration landscapes was proposed by drawing parallels between the parameter configuration space and traditional fitness landscapes, thereby facilitating the use of fitness landscape analysis in the parameter configuration domain. This paper extends the idea of the parameter configuration landscape and proposes the use of geomorphons, a physical landform classification scheme, to visualize and characterize the parameter configuration landscape. Additionally, a measure of ruggedness is proposed to quantity the difficulty associated with tuning the control parameters for an algorithm. The proposed methodology is applied to the parameter configuration landscape of the differential evolution (DE) algorithm on 20 benchmark problems in 10, 30, and 50 dimensions.

Given a set of candidate projects, selecting and scheduling an optimal subset of the projects is a complex problem faced by many organizations. This problem is referred to as the Project Portfolio Selection and Scheduling Problem (PPSSP) and is known to be NP-hard. In the defence sector, the PPSSP arises as a sub-process of Future Force Design (FFD), which is a strategic planning task that assists in the decision making process for future defence force capability programming. The PPSSP faced in the defence context has its own set of challenges above and beyond those typical of the problem. As such, this study investigates a formulation of the PPSSP inspired by the FFD process in the context of the Australian Department of Defence (DoD). Given the NP-hard nature of the problem, four metaheuristics are examined on large-scale synthetic data sets. Three different solution representations are examined and results are compared against solutions provided by a commercial exact solver. Results indicate that there is no observed significant difference in total portfolio value attained by the proposed meta-heuristic approaches and the commercial solver, thereby justifying their usage in this domain.

Future force design is a crucial task that assists in the creation of an effective future defence force. The primary objective of this task is to select a set of projects, within a fixed planning window and subject to budgetary constraints, that will lead to improved capabilities. While inherently related to the well-known multi-period knapsack problem, addressing this problem in the context of the defence sector gives rise to a number of unique nuances and associated challenges. Furthermore, the literature pertaining to the selection and scheduling of projects for capability-based planning in the defence sector is rather limited. To address this literature gap, this paper formalizes a multi-period project selection and scheduling problem inspired by future force design. Numerous heuristics, both random and deterministic, along with a hybrid genetic algorithm, are employed to optimize a set of instances of the proposed problem formulation with various characteristics derived from real-world, public defence data made available by the Australian Department of Defence.

It is well known that tuning a meta-heuristic optimizer is a challenging, yet rewarding process. Despite the benefits of a properly tuned optimizer, there is very little that is understood about the actual tuning process - many automated parameter tuning methods use an assumption that parameter configurations near a promising configuration will also be promising. However, this assumption has not been verified, in general. While the field of fitness landscape analysis can provide insight into the difficulty of an optimization problem, these techniques have not yet been applied to the parameter tuning problem. This paper proposes a methodology to apply standard techniques from fitness landscape analysis to the parameter configuration landscape of an arbitrary optimizer. This allows the characterization of the parameter tuning problem for an arbitrary optimizer on an arbitrary optimization problem. The proposed methodology is then investigated for the particle swarm optimization (PSO) algorithm on 20 benchmark problems in both 10 and 30 dimensions. The results indicate that the parameter configuration landscape of the PSO algorithm is globally unimodal, yet not necessarily an easy landscape to search. Furthermore, it is found that the characteristics of the PSO parameter configuration landscape do not correlate with the characteristics of the target benchmark problems.

Particle swarm optimization (PSO) is a stochastic search algorithm based on the social dynamics of a flock of birds. The performance of the PSO algorithm is known to be sensitive to the values assigned to its control parameters, and appropriate tuning of these control parameters can greatly improve performance. This paper employs function analysis of variance (fANOVA) to quantify the importance of each of the three conventional PSO control parameters, namely the inertia weight (¿), the cognitive acceleration coefficient (c1), and the social acceleration coefficient (c2), according to their respective variances associated with the fitness. Results indicate that the inertia value, ¿, has the greatest sensitivity to its assigned value and thus is the most important parameter to tune when optimizing PSO performance for low dimensional problems.

The performance of the particle swarm optimization (PSO) algorithm is known to be sensitive to the values of its control parameters. Parameter tuning is thus an important aspect in optimizing PSO performance. While many studies have examined a variety of PSO parameters and have provided general-purpose parameter suggestions, a recent study has shown that the best parameters to employ are, in fact, time-dependent. Furthermore, a priori parameter tuning is a time-consuming procedure and assumes that the best parameters to employ do not change over time. This study proposes a new PSO variant which randomly samples its control parameter values from a region known to contain promising parameter configurations, thereby eliminating the need to specify (and tune) values for the traditional PSO control parameters. The new PSO variant is compared to PSO employing 14 different parametrizations suggested in the literature. Results indicate that the performance of the proposed variant is on par with the best parameter configurations suggested in the literature.

This paper examines the position update equation of the particle swarm optimization (PSO) algorithm, leading to the proposal of a simplified position update based upon a Gaussian distribution. The proposed algorithm, Gaussian-valued particle swarm optimization (GVPSO), generates probabilistic positions by retaining key elements of the canonical update procedure while also removing the need to specify values for the traditional PSO control parameters. Experimental results across a set of 60 benchmark problems indicate that GVPSO outperforms both the standard PSO and the bare bones particle swarm optimization (BBPSO) algorithm, which also employs a Gaussian distribution to generate particle positions.

Particle swarm optimization (PSO) is a stochastic search algorithm based on the social dynamics of a flock of birds. The performance of the PSO algorithm is known to be sensitive to the values assigned to its control parameters. While many studies have provided reasonable ranges in which to initialize the parameters based on their long-term behaviours, such previous studies fail to quantify the empirical performance of parameter configurations across a wide variety of benchmark problems. This paper specifically address this issue by examining the performance of a set of 1012 parameter configurations of the PSO algorithm over a set of 22 benchmark problems using both the global-best and local-best topologies. Results indicate that, in general, parameter configurations which are within close proximity to the boundaries of the best-known theoretically-defined convergent region lead to better performance than configurations which are further away. Moreover, results indicate that neighbourhood topology plays a far more significant role than modality and separability when determining the regions in parameter space which perform well.

The performance of the Particle Swarm Optimization (PSO) algorithm can be greatly improved if the parameters are appropriately tuned. However, tuning the control parameters for PSO algorithms has traditionally been a time-consuming, empirical process. Furthermore, ideal parameters may be time-dependent. To address the issue of parameter tuning, self-adaptive PSO (SAPSO) algorithms adapt the PSO control parameters over time. While many such SAPSO techniques have been proposed, their behaviour is not well understood as no in-depth critical analysis of their adaptation mechanisms has been performed. This study examines the convergence behaviour of eight SAPSO algorithms both analytically and empirically. Evidence clearly indicates that the field of self-adaptive PSO algorithms is in a sad state, given that many techniques either demonstrate divergent behaviour coupled with excessive invalid particles, and thus infeasible solutions, or have prohibitively low particle step sizes caused by rapid convergence.

Complex networks are systems of entities that are interconnected through meaningful relationships, resulting in structures that have statistical complexities not formed by random chance. Many graph model algorithms have been proposed to model the observed behaviours of complex networks. However, constructing such graph models manually is both tedious and problematic. Moreover, many of the models proposed in the literature have been cited as having inaccuracies with respect to the complex networks they represent. Although recent studies have proposed using genetic programming to automate the construction of graph model algorithms, only one such study has considered directed networks. This paper proposes a GP-based inference system that automatically constructs graph models for directed complex networks. Furthermore, the system proposed in this paper facilitates the use of vertex attributes, e.g., age, to incorporate network semantics - something which previous works lack. The GP system was used to reproduce three well-known graph models. Results indicate that the networks generated by the (automatically) constructed models were structurally similar to networks generated by their respective target models.

The quantum particle swarm optimization (QPSO) algorithm is a variant of the traditional particle swarm optimization (PSO) algorithm aimed at solving dynamic optimization problems. Some particles in the QPSO algorithm are selected as 'quantum' particles and the positions of these particles are sampled, using some probability distribution, within a radius (i.e., a hypersphere) around the global best position while the remainder of particles follow standard PSO behaviour. The exploration and exploitation of the QPSO algorithm is heavily influenced by the probability distribution used as well as the size of the quantum radius. However, the best probability distribution and radius size are both problem and environment dependent. This work proposes using a parent centric crossover (PCX) operator to generate the positions of quantum particles, thereby removing the need for radius and probability distribution parameters completely. Two variants are proposed and results indicate that both variants are superior to QPSO, especially in environments exhibiting high temporal severity.