W B Langdon's 2005 Abstracts

W.B.Langdon . 21 August 2014 2005 papers , full list


Kernel methods for PSOs

W. B. Langdon and Riccardo Poli and Christopher R. Stephens. Technical report CSM-443, ISSN 1744-8050, Dec 2005, Essex University.

ABSTRACT

We report initial experiments on evolving kernel functions which describe the average behaviour of a swarm of particles as if it was responding as a single point moving on a landscape transformed by the kernel. Such kernels may help explain swarm and population approaches leading to extended population systems, XPS. The swarm of particles are from a simple particle swarm optimiser solving one dimensional multi-modal 3 peaks and Rastrigin problems. The standard Java genetic programming implementation TinyGP is used.

Bibliographic details


Evolutionary Solo Pong Players

W. B. Langdon and Riccardo Poli in Proceedings of the 2005 IEEE Congress on Evolutionary Computation, CEC 2005, volume 3, pp2621-2628, 2-5 September, Edinburgh, UK. PDF ps.gz. (Technical report CSM-423, ISSN 1744-8050, Essex University).

ABSTRACT

An Internet Java Applet allows users anywhere to play the Solo Pong game. We compare people's performance to a hand coded ``Optimal'' player and programs automatically produced by artificial intelligence. The AI techniques are: genetic programming, including a hybrid of GP and a human designed algorithm, and a particle swarm optimiser. The AI approaches are not fine tuned. GP and PSO find good players. Evolutionary computation (EC) is able to beat both human designed code and human players.

Bibliographic details


Evolving Problems to Learn about Particle Swarm and other Optimisers

W. B. Langdon and Riccardo Poli in Proceedings of the 2005 IEEE Congress on Evolutionary Computation, CEC 2005, volume 1, pp81-88, 2-5 September, Edinburgh, UK. PDF ps.gz (slides).
Animations, benchmarks and code.

Two page version (PDF ps.gz) presented at BNAIC 2005 pp 365-366.

ABSTRACT

We use evolutionary computation (EC) to automatically find problems which demonstrate the strength and weaknesses of modern search heuristics. In particular we analyse Particle Swarm Optimization (PSO) and Differential Evolution (DE). Both evolutionary algorithms are contrasted with a robust deterministic gradient based searcher (based on Newton-Raphson). The fitness landscapes made by genetic programming (GP) are used to illustrate difficulties in GAs and PSOs thereby explaining how they work and allowing us to devise better extended particle swarm systems (XPS).

Bibliographic details


The Distribution of Amorphous Computer Outputs

W. B. Langdon. Position paper (eprint) at The Grand Challenge in Non-Classical Computation: International Workshop, Susan Stepney and Stephen Emmott Editors, 18-19 April 2005, York, UK.

ABSTRACT

Fitness distributions (landscapes) of programs tend to a limit as they get bigger. Markov minorization gives upper bounds ((15.3 + 2.30 m)/log(I)) on the length of program run on random or average computing devices. I is the size of the instruction set and m size of output register. Almost all programs are constants. Convergence is exponential with 90% of programs of length 1.6 n 2**N yielding constants (n=size input register and size of memory=N). This is supported by experiment.

Bibliographic details


Evolutionary Solo Pong Players

W. B. Langdon and Riccardo Poli Technical report CSM-423, ISSN 1744-8050, Essex University

ABSTRACT

An Internet Java Applet allows users anywhere to play the Solo Pong game. We compare people's performance to a hand coded ``Optimal'' player and programs automatically produced by artificial intelligence. The AI techniques are: genetic programming, including a hybrid of GP and a human designed algorithm, and a particle swarm optimiser. The AI approaches are not fine tuned. GP and PSO find good players. Evolutionary computation (EC) is able to beat both human designed code and human players.

Bibliographic details


Understanding Particle Swarm Optimisation by Evolving Problem Landscapes

W. B. Langdon Riccardo Poli, Owen Holland and Thiemo Krink (PDF) (gzip ps) to be presented at Swarm Intelligence Symposium 2005 SIG 05, Luca Maria Gambardella, Payman Arabshahi and Alcherio Martinoli (editors), pp30-37, 8-10 June 2005, Pasadena, California, USA.

Two animated slides: constriction helps, hinders.

ABSTRACT

Genetic programming (GP) is used to create fitness landscapes which highlight strengths and weaknesses of different types of PSO and to contrast population-based swarm approaches with non stochastic gradient followers (i.e. hill climbers). These automatically generated benchmark problems yield insights into the operation of PSOs, illustrate benefits and drawbacks of different population sizes and constriction (friction) coefficients, and reveal new swarm phenomena such as deception and the exploration/exploitation tradeoff. The method could be applied to any type of optimizer.

Bibliographic details


Pfeiffer - A Distributed Open-ended Evolutionary System

W. B. Langdon, in AISB'05: Proceedings of the Joint Symposium on Socially Inspired Computing, (METAS 2005) Bruce Edmonds, Nigel Gilbert, Steven Gustafson, David Hales and Natalio Krasnogor Editors, 12-15 April 2005, University of Hertfordshire, Hatfield, England, pp. 7-13. (PDF).

ABSTRACT

Pfeiffer contains a population of fractals which has been evolving continuously for more than three years. The animations are developed from embryos using a Lindenmayer grammar (L-System). These open generative representations potentially allow gene duplication and the evolution of higher order genetic operators and might be a step towards the emergence of social intelligence in swarms of artificial life (alife) agents. The fitness function is simply do the snowflake patterns appeal to the users: interactive evolution (IEC). To this end, images are placed in animated snow globes (computerised snowstorms) by world wide web (www) browsers (Netscape, Mozilla, Internet Explorer, Firefox, etc.) anywhere on the planet. More than 600 people have used http://www.cs.ucl.ac.uk/staff/W.Langdon/pfeiffer.html.

Bibliographic details


Repeated Patterns in Tree Genetic Programming

W. B. Langdon and W. Banzhaf, (PDF gzipped postscript), Slides presented at EuroGP-2005. LNCS 3447, 30 March - 1 April 2005 Lausanne, p190-202 doi:10.1007/b107383

ABSTRACT

We extend our analysis of repetitive patterns found in genetic programming genomes * to tree based GP. As in linear GP, repetitive patterns are present in large numbers. Size fair crossover limits bloat in automatic programming, preventing the evolution of recurring motifs. We examine these complex properties in detail: e.g. using depth v. size Catalan binary tree shape plots, subgraph and subtree matching, information entropy, syntactic and semantic fitness correlations and diffuse introns. We relate this emergent phenomenon to considerations about building blocks in GP and how GP works.

Bibliographic details


Genetic Programming and Evolvable Machines: Five years of Reviews

W.B. Langdon and S. Gustafson. (PDF gzipped postscript), Genetic Programming and Evolvable Machines, 2005, 6(2) pp.221-228. doi:10.1007/s10710-005-6165-9

ABSTRACT

The journal, and in particular the resource reviews have been running for more than five years. Now is a good time to revisit our original goals compare them with what the journal has achieved and make new plans. Section 2 onwards, updates the statistics on the state of the genetic programming, evolvable hardware and evolvable machines literature and electronic resources.

Bibliographic details


Repeated Sequences in Linear Genetic Programming Genomes

William B. Langdon and Wolfgang Banzhaf, Complex Systems 15 (4) pp285-306. (PDF gzipped postscript). Extends GECCO'2004 late breaking paper.

ABSTRACT

Biological chromosomes are replete with repetitive sequences, microsatellites, SSR tracts, ALU, etc. in their DNA base sequences. We started looking for similar phenomena in evolutionary computation. First studies find copious repeated sequences, which can be hierarchically decomposed into shorter sequences, in programs evolved using both homologous and two point crossover but not with headless chicken crossover or other mutations. In bloated programs the small number of effective or expressed instructions appear in both repeated and non-repeated code. Hinting that building-blocks or code reuse may evolve in unplanned ways.

Mackey-Glass chaotic time series prediction and eukaryotic protein localisation (both previously used as artificial intelligence machine learning benchmarks) demonstrate evolution of Shannon information (entropy) and lead to models capable of lossy Kolmogorov compression. Our findings with diverse benchmarks and GP systems suggest this emergent phenomenon may be widespread in genetic systems.

Data set Bibliographic details


up
W.B.Langdon cs.ucl.ac.uk