Analysing the evolvability of neural network agents through structural mutations

Authors:
Ehud Schlessinger;Peter J. Bentley;R. Beau Lotto
Affiliations:
Institute of Ophthalmology, University College London, London;Department of Computer Science, University College London, London;Institute of Ophthalmology, University College London, London
Venue:
ECAL'05 Proceedings of the 8th European conference on Advances in Artificial Life
Year:
2005

Citing 7
Cited 4

The evolution of evolvability in genetic programming

Advances in genetic programming
How neutral networks influence evolvability

Complexity
Fitness landscapes and evolvability

Evolutionary Computation
Evolving neural networks through augmenting topologies

Evolutionary Computation
Genome Growth and the Evolution of the Genotype-Phenotype Map

Evolution and Biocomputation, Computational Models of Evolution
Evolving visually guided agents in an ambiguous virtual world

GECCO '05 Proceedings of the 7th annual conference on Genetic and evolutionary computation
Design of evolvable computer languages

IEEE Transactions on Evolutionary Computation

Modular thinking: evolving modular neural networks for visual guidance of agents

Proceedings of the 8th annual conference on Genetic and evolutionary computation
On the effects of node duplication and connection-oriented constructivism in neural XCSF

Proceedings of the 10th annual conference companion on Genetic and evolutionary computation
Towards continuous actions in continuous space and time using self-adaptive constructivism in neural XCSF

Proceedings of the 11th Annual conference on Genetic and evolutionary computation
Phased searching with NEAT in a Time-Scaled Framework: Experiments on a computer-aided detection system for lung nodules

Artificial Intelligence in Medicine

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper investigates evolvability of artificial neural networks within an artificial life environment. Five different structural mutations are investigated, including adaptive evolution, structure duplication, and incremental changes. The total evolvability indicator, Etotal, and the evolvability function through time, are calculated in each instance, in addition to other functional attributes of the system. The results indicate that incremental modifications to networks, and incorporating an adaptive element into the evolution process itself, significantly increases neural network evolvability within open-ended artificial life simulations.