arXiv subjects

Larry Bull

Publications and source records attributed to Larry Bull.

At least 55 records · Page 3Linked to original sources

arXiv2014

Discrete Dynamical Genetic Programming in XCS

A number of representation schemes have been presented for use within Learning Classifier Systems, ranging from binary encodings to neural networks. This paper presents results from an investigation into using a discrete dynamical system representation within the XCS Learning Classifier System. In particular, asynchronous random Boolean networks are used to represent the traditional condition-action production system rules. It is shown possible to use self-adaptive, open-ended evolution to design an ensemble of such discrete dynamical systems within XCS to solve a number of well-known test problems.

cs.AI↗

arXiv2014

A Brief History of Learning Classifier Systems: From CS-1 to XCS

Modern Learning Classifier Systems can be characterized by their use of rule accuracy as the utility metric for the search algorithm(s) discovering useful rules. Such searching typically takes place within the restricted space of co-active rules for efficiency. This paper gives an historical overview of the evolution of such systems up to XCS, and then some of the subsequent developments of XCS to different types of learning.

cs.NE↗

arXiv2014

Exploiting generalisation symmetries in accuracy-based learning classifier systems: An initial study

Modern learning classifier systems typically exploit a niched genetic algorithm to facilitate rule discovery. When used for reinforcement learning, such rules represent generalisations over the state-action-reward space. Whilst encouraging maximal generality, the niching can potentially hinder the formation of generalisations in the state space which are symmetrical, or very similar, over different actions. This paper introduces the use of rules which contain multiple actions, maintaining accuracy and reward metrics for each action. It is shown that problem symmetries can be exploited, improving performance, whilst not degrading performance when symmetries are reduced.

cs.NE↗

arXiv2013

Towards Application of the RBNK Model

The computational modeling of genetic regulatory networks is now common place, either by fitting a system to experimental data or by exploring the behaviour of abstract systems with the aim of identifying underlying principles. This paper presents an approach to the latter, considering the response to environmental changes of a well-known model placed upon tunable fitness landscapes. The effects on genome size and gene connectivity are explored.

cs.CE↗

arXiv2013

Evolving Gene Regulatory Networks with Mobile DNA Mechanisms

This paper uses a recently presented abstract, tuneable Boolean regulatory network model extended to consider aspects of mobile DNA, such as transposons. The significant role of mobile DNA in the evolution of natural systems is becoming increasingly clear. This paper shows how dynamically controlling network node connectivity and function via transposon-inspired mechanisms can be selected for in computational intelligence tasks to give improved performance. The designs of dynamical networks intended for implementation within the slime mould Physarum polycephalum and for the distributed control of a smart surface are considered.

cs.CE↗

arXiv2013

Evolving Boolean Regulatory Networks with Epigenetic Control

The significant role of epigenetic mechanisms within natural systems has become increasingly clear. This paper uses a recently presented abstract, tunable Boolean genetic regulatory network model to explore aspects of epigenetics. It is shown how dynamically controlling transcription via a DNA methylation-inspired mechanism can be selected for by simulated evolution under various single and multiple cell scenarios. Further, it is shown that the effects of such control can be inherited without detriment to fitness.

cs.NE↗

arXiv2013

On Mobile DNA in Artificial Regulatory Networks: Evolving Functional and Structural Dynamism

There is a growing body of work considering the use of representations based upon genetic regulatory networks. This paper uses a recently presented abstract, tunable Boolean regulatory network model to explore aspects of mobile DNA, such as transposons, within these dynamical systems. The significant role of mobile DNA in the evolution of natural systems is becoming increasingly clear. Whilst operators loosely based upon transposons have previously been used within evolutionary computation, their use within regulatory network representations enables the potential exploitation of numerous new mechanisms. This paper shows how dynamically controlling network node connectivity and function via transposon-inspired mechanisms can be selected for under non-stationary and coevolutionary scenarios, including when such changes are heritable.

cs.ET↗

arXiv2012

Towards the Evolution of Novel Vertical-Axis Wind Turbines

Renewable and sustainable energy is one of the most important challenges currently facing mankind. Wind has made an increasing contribution to the world's energy supply mix, but still remains a long way from reaching its full potential. In this paper, we investigate the use of artificial evolution to design vertical-axis wind turbine prototypes that are physically instantiated and evaluated under approximated wind tunnel conditions. An artificial neural network is used as a surrogate model to assist learning and found to reduce the number of fabrications required to reach a higher aerodynamic efficiency, resulting in an important cost reduction. Unlike in other approaches, such as computational fluid dynamics simulations, no mathematical formulations are used and no model assumptions are made.

cs.NE↗

arXiv2012

Evolution of Plastic Learning in Spiking Networks via Memristive Connections

This article presents a spiking neuroevolutionary system which implements memristors as plastic connections, i.e. whose weights can vary during a trial. The evolutionary design process exploits parameter self-adaptation and variable topologies, allowing the number of neurons, connection weights, and inter-neural connectivity pattern to emerge. By comparing two phenomenological real-world memristor implementations with networks comprised of (i) linear resistors (ii) constant-valued connections, we demonstrate that this approach allows the evolution of networks of appropriate complexity to emerge whilst exploiting the memristive properties of the connections to reduce learning time. We extend this approach to allow for heterogeneous mixtures of memristors within the networks; our approach provides an in-depth analysis of network structure. Our networks are evaluated on simulated robotic navigation tasks; results demonstrate that memristive plasticity enables higher performance than constant-weighted connections in both static and dynamic reward scenarios, and that mixtures of memristive elements provide performance advantages when compared to homogeneous memristive networks.

cs.ET↗

arXiv2012

Coevolving Cellular Automata with Memory for Chemical Computing: Boolean Logic Gates in the B-Z Reaction

We propose that the behaviour of non-linear media can be controlled automatically through coevolutionary systems. By extension, forms of unconventional computing, i.e., massively parallel non-linear computers, can be realised by such an approach. In this study a light-sensitive sub-excitable Belousov-Zhabotinsky reaction is controlled using various heterogeneous cellular automata. A checkerboard image comprising of varying light intensity cells is projected onto the surface of a catalyst-loaded gel resulting in rich spatio-temporal chemical wave behaviour. The coevolved cellular automata are shown to be able to control chemical activity through dynamic control of the light intensity. The approach is demonstrated through the creation of a number of simple Boolean logic gates.

cs.ET↗

arXiv2012

Initiation of waves in BZ encapsulated vesicles using light - towards design of computing architectures

A gas free analogue of the Belousov-Zhabotinsky reaction catalysed by ferroin and encapsulated in phospholipid stabilised vesicles is reported. A reaction mixture which exhibits spontaneous oscillation and excitation transfer between vesicles was formulated. By adjusting the reagent concentrations a quiescent state with fewer spontaneous oscillations was achieved. Using relatively low power laser sources of specific wavelengths (green 532nm and blue 405nm) it was shown that waves could be reproducibly initiated within the BZ vesicles. Furthermore, despite the reduced excitability of the system overall the initiated waves exhibited vesicle to vesicle transfer. It was possible to manipulate single vesicles and design simple circuits based on a 2D validation of collision based circuits. Therefore, we conclude that this BZ system exhibits promise for computing applications based on 3D networks of vesicles.

nlin.PS↗

arXiv2012

Toward Turing's A-type Unorganised Machines in an Unconventional Substrate: a Dynamic Representation in Compartmentalised Excitable Chemical Media

Turing presented a general representation scheme by which to achieve artificial intelligence - unorganised machines. Significantly, these were a form of discrete dynamical system and yet such representations remain relatively unexplored. Further, at the same time as also suggesting that natural evolution may provide inspiration for search mechanisms to design machines, he noted that mechanisms inspired by the social aspects of learning may prove useful. This paper presents initial results from consideration of using Turing's dynamical representation within an unconventional substrate - networks of Belousov-Zhabotinsky vesicles - designed by an imitation-based, i.e., cultural, approach. Turing's representation scheme is also extended to include a fuller set of Boolean functions at the nodes of the recurrent networks.

cs.ET↗

arXiv2012

Fuzzy Dynamical Genetic Programming in XCSF

A number of representation schemes have been presented for use within Learning Classifier Systems, ranging from binary encodings to Neural Networks, and more recently Dynamical Genetic Programming (DGP). This paper presents results from an investigation into using a fuzzy DGP representation within the XCSF Learning Classifier System. In particular, asynchronous Fuzzy Logic Networks are used to represent the traditional condition-action production system rules. It is shown possible to use self-adaptive, open-ended evolution to design an ensemble of such fuzzy dynamical systems within XCSF to solve several well-known continuous-valued test problems.

cs.AI↗

arXiv2012

Production System Rules as Protein Complexes from Genetic Regulatory Networks

This short paper introduces a new way by which to design production system rules. An indirect encoding scheme is presented which views such rules as protein complexes produced by the temporal behaviour of an artificial genetic regulatory network. This initial study begins by using a simple Boolean regulatory network to produce traditional ternary-encoded rules before moving to a fuzzy variant to produce real-valued rules. Competitive performance is shown with related genetic regulatory networks and rule-based systems on benchmark problems.

cs.NE↗

arXiv2012

A Spiking Neural Learning Classifier System

Learning Classifier Systems (LCS) are population-based reinforcement learners used in a wide variety of applications. This paper presents a LCS where each traditional rule is represented by a spiking neural network, a type of network with dynamic internal state. We employ a constructivist model of growth of both neurons and dendrites that realise flexible learning by evolving structures of sufficient complexity to solve a well-known problem involving continuous, real-valued inputs. Additionally, we extend the system to enable temporal state decomposition. By allowing our LCS to chain together sequences of heterogeneous actions into macro-actions, it is shown to perform optimally in a problem where traditional methods can fail to find a solution in a reasonable amount of time. Our final system is tested on a simulated robotics platform.

cs.NE↗

arXiv2012

On Natural Genetic Engineering: Structural Dynamism in Random Boolean Networks

This short paper presents an abstract, tunable model of genomic structural change within the cell lifecycle and explores its use with simulated evolution. A well-known Boolean model of genetic regulatory networks is extended to include changes in node connectivity based upon the current cell state, e.g., via transposable elements. The underlying behaviour of the resulting dynamical networks is investigated before their evolvability is explored using a version of the NK model of fitness landscapes. Structural dynamism is found to be selected for in non-stationary environments and subsequently shown capable of providing a mechanism for evolutionary innovation when such reorganizations are inherited.

cs.CE↗

arXiv2011

Logical and Arithmetic Circuits in Belousov Zhabotinsky Encapsulated Discs

Excitation waves on a sub-excitable Belousov Zhabotinsky (BZ) substrate can be manipulated by chemical variations in the substrate and by interactions with other waves. Symbolic assignment and interpretation of wave dynamics can be used to perform logical and arithmetic computations. We present chemical analogs of elementary logic and arithmetic circuits created entirely from interconnected arrangements of individual BZ encapsulated cell like discs. Inter-disc wave migration is confined in carefully positioned connecting pores. This connection limits wave expansion and unifies the input-output characteristic of the discs. Circuit designs derived from numeric simulations are optically encoded onto a homogeneous photo-sensitive BZ substrate.

nlin.PS↗

arXiv2011

Vesicle computers: Approximating Voronoi diagram on Voronoi automata

Irregular arrangements of vesicles filled with excitable and precipitating chemical systems are imitated by Voronoi automata --- finite-state machines defined on a planar Voronoi diagram. Every Voronoi cell takes four states: resting, excited, refractory and precipitate. A resting cell excites if it has at least one excited neighbour; the cell precipitates if a ratio of excited cells in its neighbourhood to its number of neighbours exceed certain threshold. To approximate a Voronoi diagram on Voronoi automata we project a planar set onto automaton lattice, thus cells corresponding to data-points are excited. Excitation waves propagate across the Voronoi automaton, interact with each other and form precipitate in result of the interaction. Configuration of precipitate represents edges of approximated Voronoi diagram. We discover relation between quality of Voronoi diagram approximation and precipitation threshold, and demonstrate feasibility of our model in approximation Voronoi diagram of arbitrary-shaped objects and a skeleton of a planar shape.

cs.ET↗