Publications

Understanding the origins of complexity is a key challenge in many sciences. Although networks are known to underlie most systems, showing how they contribute to well-known phenomena remains an issue. Here, we show that recurrent phase transitions in network connectivity underlie emergent phenomena in many systems. We identify properties that are typical of systems in different connectivity phases, as well as characteristics commonly associated with the phase transitions. We synthesize these common features into a common framework, which we term dual-phase evolution (DPE). Using this framework, we review the literature from several disciplines to show that recurrent connectivity phase transitions underlie the complex properties of many biological, physical and human systems. We argue that the DPE framework helps to explain many complex phenomena, including perpetual novelty, modularity, scale-free networks and criticality. Our review concludes with a discussion of the way DPE relates to other frameworks, in particular, self-organized criticality and the adaptive cycle.

This is a republication of this 2009 paper.

Evidence from several fields suggests that dual phase evolution (DPE) may account for distinctive features associated with complex adaptive systems. Here, we review empirical and theoretical evidence for DPE in natural systems and examine the relationship of DPE to self-organized criticality and adaptive cycles. A general model for DPE in networks is outlined, with preliminary data illustrating the emergence of phase changes.

This is a republication of this 2009 paper.

Data show that human-like cognitive traits do not evolve in animals through natural selection. Rather, human-like cognition evolves through runaway selection for social skills. Here, we discuss why social selection may be uniquely effective for promoting human-like cognition, and the conditions that facilitate it. These observations suggest future directions for artificial life research aimed at generating human-like cognition in digital organisms.

A central question in complexity theory is how large-scale phenomena, such as such as self-organisation, perpetual novelty, and sustained diversity, emerge. Complex systems can be understood as networks of interacting components. The focus of this research is the role that the properties of such networks play in self-organisation and emergence in complex systems.

Based on the previously known concept of Dual Phase Evolution (DPE), I propose a theoretical framework, within which recurrent phase transitions in network connectivity underlie emergent phenomena in many systems. This DPE framework extends and refines the original concept.

Networks can exist in two general connectivity phases: well connected and poorly connected. DPE relates each of the two connectivity phases and the transition events between them to typical system dynamics. I analyse empirical and experimental evidence from published studies in areas as diverse as physics, biology, socio-economics, mathematics and computer science. The analysis implies that DPE is widespread and operates in many kinds of complex systems, where it drives emergence and self-organisation. What is more, the analysis uncovers hitherto unstudied deep similarities and common underlying processes between different complex systems.

To further understand the theoretical concepts of the DPE framework, I apply DPE in studies of mechanisms behind particular emergent properties in several types of complex systems:

Seeking to better understand the emergence of novelty and diversity in ecosystems, I develop and study an individual-based simulation model of adaptive radiation (speciation) in landscapes. Simulation results imply that recurrent external disturbances facilitate perpetual novelty and diversity in landscape populations through two complementary mechanisms: One mechanism constitutes recurrent DPE phase changes in landscape connectivity on several levels. The other mechanism is alteration of the environment in disturbed areas leading to modified selection regimes.

As a result of the simulation studies of landscape evolution, I develop a new genetic model that combines the advantages of two existing genetic models. The new model allows individual-based simulation studies of genetics on holey fineness landscapes (HFLs). Such fitness landscapes result from biochemical constraints to genetic viability and have previously only been studied analytically. Simulation studies of reproductive isolation uncover that when HFLs are considered, common predictions about maintenance of reproductive isolation in migrating populations change. Results also show that HFL-genetics can facilitate the emergence of stable hybrid populations, and the evolution of social selection though reinforcement.

Continuing to study and apply DPE, I investigate how DPE processes can lead to the emergence of important network topologies. Using simulations models, I demonstrate two possible mechanisms behind emergent connectivity phase transitions without facilitation by external stimuli. A study of social network models reveals simple mechanisms that lead to structures typical of some real social networks and points towards general principles for emergence of important topologies such as modularity. A study of a network model of co-operations in markets reveals further mechanisms behind the emergence of complex and hierarchical modularity.

Generative models for scale-free networks, that are ubiquitous in many natural systems, are well known, however, such models apply to growing networks. I propose and examine a generative model for scale-free topologies that can account for some scale-free networks of constant size found in nature.

A wider context for DPE as a framework for reasoning about complexity is provided by examining the relationship between DPE and other established concepts such as Self-Organised Criticality and the Adaptive Cycle. In conclusion, DPE complements other established theories.

In general, network-theoretical approaches, such as DPE, are powerful paradigms in understanding complexity. This thesis shows that recurrent changes in connectivity of component interaction networks constitute a broad mechanism for emergence and self-organisation in complex systems, and demonstrates this mechanism in several specific biological and socio-economic systems.

Data show that human-like cognitive traits do not evolve in animals through natural selection. Rather, human-like cognition evolves through runaway selection for social skills. Here, we discuss why social selection may be uniquely effective for promoting human-like cognition, and the conditions that facilitate it. These observations suggest future directions for artificial life research aimed at generating human-like cognition in digital organisms.

Evidence from several fields suggests that Dual Phase Evolution (DPE) may account for distinctive features associated with complex adaptive systems. We review empirical and theoretical evidence for DPE in natural systems and examine the relationship of DPE to other complexity theories such self-organised criticality and adaptive cycles.

Evidence from several fields suggests that dual phase evolution (DPE) may account for distinctive features associated with complex adaptive systems. Here, we review empirical and theoretical evidence for DPE in natural systems and examine the relationship of DPE to self-organized criticality and adaptive cycles. A general model for DPE in networks is outlined, with preliminary data illustrating the emergence of phase changes.

Dual Phase Evolution (DPE) is a widespread natural process in which complex systems adapt and self-organise by switching alternately between two phases: a phase of global interactions and a phase of local interactions. We show that in evolving networks of agents, DPE can give rise to a wide variety of topologies. In particular, it can lead to the spontaneous emergence of stabilising modular structure.

Complex adaptive and evolutionary systems exhibit a sustained diversity, far-from-equilibrium dynamics, and permanent novelty and adaptation in the absence of a global controller. Previous work shows that many such systems can be represented as networks of interacting components. These networks are typified by certain complex topologies. Insights into the processes behind the emergence of complex network structures and into the effects of such structures are necessary for an understanding of properties that characterise adaptive and evolutionary systems.

Dual Phase Evolution (DPE) is a widespread natural process in which networks underlying complex systems adapt and self-organise by switching alternately between two phases: a phase of global interactions and a phase of local interactions. Each phase is characterised by specific global connectivity and interaction patterns. In this talk I present ongoing work on DPE in complex evolutionary systems. I show how DPE processes can give rise to a wide variety of complex network topologies. In particular, this includes the emergence of scale-free degree distributions fixed-size networks, as well as modular structures. I also show how DPE can be responsible to the continuous novelty observed in many natural and artificial evolutionary systems.

This talk is based on a paper about DPE and Self-Organisation in Networks that was published by myself and my colleagues at SEAL'08.

For further information, please refer to a list of DPE-related papers that I co-authored and follow the references.

A genetic algorithm (GA) is utilised to discover known and novel PROSITE-like sequence templates that can be used to classify the sub-cellular location of eukaryotic proteins. While traditional machine learning techniques present a black-box approach to this problem, the current method explicitly represents the discovered localisation motifs. A combined multi-class location classifier is presented and compared to other techniques based on genetic programming. Without consideration of additional structural information the presented method outperforms the alternative techniques.

Analytical models show that high-dimensional fitness landscapes form “holey” rather than “rugged” topographies, but the implications of this finding for biological and artificial life systems remain largely unexplored. One of the reasons for this gap can be attributed to serious difficulties in the implementation of individual-based holey fitness landscape (HFL) models. Here, we introduce a method for simulating HFLs in spatially explicit individual-based models that overcomes these difficulties. We examine how the HFL changes predictions for the maintenance of genetic diversity in the face of migration. Previous models suggest that ecologically-based reproductive isolation will rapidly collapse under migration. Our results indicate that an underlying HFL can often maintain diversity in this situation. Hybrid species emerge frequently when HFL genetics are simulated, but are usually doomed to extinction because of small population sizes. However, hybridisation can also lead to novel adaptations and potentially the exploitation of new ecological niches. More generally, the results imply that HFL genetics should not be neglected in studies of adaptation and diversity.

During the last three decades, the field of Intelligent and Evolutionary Systems (IES) has established itself as a key research area, with the ability to address common problems across diverse disciplines. Many challenges arise from the massive volumes of data provided by widespread automation, and from the increasingly complex webs of interaction that arise as local events scale up to generate global phenomena. Inspiration from biology – the core concept of IES research - underlies key advances in engineering systems that respond with adaptive intelligence and resilience in such noisy, dynamic environments.

Today, IES continues to develop, as traditional neural and genetic computation techniques are supplemented with new ideas from a diverse range of areas that include swarm computing, artificial life, complex systems and many others. Advances on key theoretical fronts are complemented by close ties to applications within specific areas, ensuring that IES retain relevance across disciplines.

The Asia Pacific Symposium on Intelligent and Evolutionary Systems is an annual event that brings together researchers from the Asia-Pacific Rim and other areas to discuss emerging ideas as well as on-going research in intelligent and evolutionary computation, and to deepen and extend links and collaboration in the community.

The 2008 event is the 12th meeting in a series of IES symposia that originated in 1997. This year, the event was held in Melbourne, Australia in conjunction with the 7th International Conference on Simulated Evolution and Learning (SEAL'08). The work presented at the symposium is indicative of the diverse areas touched by IES research, with applications in architecture, horticulture, epidemiology, software development and traffic control among others. In addition to exploring several new applications of genetic algorithms, papers here develop novel approaches to data assimilation, examine the theory and application of learning systems, and explore the evolution of culture and cooperation in multi-agent simulations. Key ideas such as non-dominance and entropy evolution are also further developed.

We are very grateful to the many people who have helped to make this year's symposium a success. First of all we thank the authors for their work and participation, and for providing manuscripts on time and in a standard format. All papers were carefully reviewed by at least two independent members of the program committee to ensure quality and consistency, and the time and effort contributed by the program committee are much appreciated. We are particularly grateful to Prof. David Green whose experience from previous IES meetings was invaluable for ensuring the success of this year's event. We also thank Ms Dianne Nguyen for her assistance with the event coordination and the production of these proceedings.

Premature convergence is a persisting problem in evolutionary optimisation, in particular – genetic algorithms. While a number of methods exist to approach this issue, they usually require problem specific calibration or only partially resolve the issue, at best by delaying the premature convergence of an evolving population. Analytical models in biology show that resiliently diverse populations evolve on high-dimensional fitness landscapes with “holey” rather than “rugged” topographies, but the implications for artificial evolutionary systems remain largely unexplored. Here I show how holey fitness landscapes (HFLs) can be incorporated in an evolutionary algorithm and use this approach to investigate the ability of HFLs to maintain genetic diversity in an evolving population. The results indicate that an underlying HFL can counteract premature genetic convergence and sustain diversity. They also suggest that HFL may provide a flexible mechanism for dynamic creation and maintenance of subpopulations that concentrate their evolutionary search in different regions of the solution space. Finally, I discuss on-going work on using the HFL model in optimisation problems.

In this study, we describe an evolutionary mechanism – Dual Phase Evolution (DPE) – and argue that it plays a key role in the emergence of internal structure in complex adaptive systems (CAS). Our DPE theory proposes that CAS exhibit two well-defined phases – selection and variation – and that shifts from one phase to the other are triggered by external perturbations. We discuss empirical data which demonstrates that DPE processes play a prominent role in species evolution within landscapes and argue that processes governing a wide range of self-organising phenomena are similar in nature. In support, we present a simulation model of adaptive radiation in landscapes. In the model, organisms normally exist within a connected landscape in which selection maintains them in a stable state. Intermittent disturbances (such as fires, commentary impacts) flip the system into a disconnected phase, in which populations become fragmented, freeing up areas of empty space in which selection pressure lessens and genetic variation predominates. The simulation results show that the DPE mechanism may indeed facilitate the appearance of complex diversity in a landscape ecosystem.

Recent studies suggest that macro-evolutionary patterns such as punctuated equilibrium may be generated by a process termed Dual Phase Evolution (DPE). According to the DPE hypothesis, evolution in landscapes exhibits two phases – selection and variation. Disturbances such as mass extinctions can flip the landscape from selection to variation phases. Similar processes occur in a wide range of artificial, natural and social complex systems. Here, we show that mass extinctions induce DPE in a simulation model of adaptive radiation. The model is based on a previous model of adaptive radiation which did not incorporate Dual Phase Evolution. Results confirm that mass extinctions caused by external disturbances can trigger periods of rapid species turnover and adaptive radiation (variation phases), which are followed by long periods without innovation (selection phases). Our simulations also show that the spatial configuration of disasters leading to mass extinctions strongly influences whether and to what extent such disasters are capable of inducing evolutionary variation phases.

We describe a process, Dual Phase Evolution (DPE), which is a widespread mechanism by which systems evolve. The process occurs in systems that switch between two phases: a balance phase and a variation phase. In DPE, processes drive systems to settle in a “balance” phase, but external stimuli can flip the system into a temporary “variation” phase, in which rapid change occurs on all scales. The system gradually returns to a new balance phase, often with different structure than formerly. We argue that the system occurs in many natural systems and show how it results from landscape connectivity in both evolution and long-term forest change. We describe how it underlies certain optimization algorithms, and can be implemented to improve the performance of existing optimization methods. Finally, we present simulation experiments that show how DPE may play a role in creating cliques, clusters, modules and other kinds of order within social networks.

We provide a brief overview of the use of FLs in evolutionary biology and introduce an FL model suitable for individual-based models of species evolution. Our model combines different features of several analytic FL models used in evolutionary biology. Our new model overcomes several difficulties encountered with previous FL models, particularly arbitrary divergence thresholds. We discuss the difficulties encountered during the implementation of our model and how we have overcome these. We present a detailed numerical analysis of the proposed family of FLs that will inform future work based on our FL model.

In this article we aim to analyse some of the advances in security of communication since this discipline evolved and to pinpoint the main problems. We then introduce a modern attempt to solve some of these problems, in particular the key distribution problem, by using the theory of quantum mechanics to construct a communication system that automatically detects eavesdropping. We examine some of the implications of quantum mechanics relevant to this field and then introduce a selection of communication protocols based on them. Finally we examine how secure these protocols are and identify their potential weaknesses.

A primary motivation for research in digital ecosystems is the desire to exploit the self-organising properties of natural ecosystems. Ecosystems are thought to be robust, scalable architectures that can automatically solve complex, dynamic problems. However, the biological processes that contribute to these properties have not been made explicit in digital ecosystem research. Here, we discuss how biological properties contribute to the self-organising features of natural ecosystems. These properties include populations of evolving agents, a complex dynamic environment, and spatial distributions which generate local interactions. The potential for exploiting these properties in artificial systems is then considered. An example architecture, the Digital Business Ecosystem (DBE), is considered in detail. Simulation results imply that the DBE performs better at large scales than a comparable service-oriented architecture. These results suggest that incorporating ideas from theoretical ecology can contribute to useful self-organising properties in digital ecosystems.