Automatic task decomposition for the neuroevolution of augmenting topologies (NEAT) algorithm

Authors:
Timmy Manning;Paul Walsh
Affiliations:
Cork Institute of Technology, Cork, Ireland;Cork Institute of Technology, Cork, Ireland
Venue:
EvoBIO'12 Proceedings of the 10th European conference on Evolutionary Computation, Machine Learning and Data Mining in Bioinformatics
Year:
2012

Citing 10
Cited 1

1994 Special Issue: Design and evolution of modular neural network architectures

Neural Networks - Special issue: models of neurodynamics and behavior
Global Optimization for Neural Network Training

Computer - Special issue: neural computing: companion issue to Spring 1996 IEEE Computational Science & Engineering
Crossover; Concepts and Applications in Genetics, Evolution, and Breeding

Crossover; Concepts and Applications in Genetics, Evolution, and Breeding
Evolving neural networks through augmenting topologies

Evolutionary Computation
A Neural Model for Context-dependent Sequence Learning

Neural Processing Letters
Adaptive mixtures of local experts

Neural Computation
Competitive coevolution through evolutionary complexification

Journal of Artificial Intelligence Research
Evolving modular neural-networks through exaptation

CEC'09 Proceedings of the Eleventh conference on Congress on Evolutionary Computation
Theoretical convergence guarantees for cooperative coevolutionary algorithms

Evolutionary Computation
Evolutionary ensembles with negative correlation learning

IEEE Transactions on Evolutionary Computation

Improving the performance of CGPANN for breast cancer diagnosis using crossover and radial basis functions

EvoBIO'13 Proceedings of the 11th European conference on Evolutionary Computation, Machine Learning and Data Mining in Bioinformatics

Quantified Score

Hi-index	0.00

Visualization

Abstract

Neuroevolution, the process of creating artificial neural networks through simulated evolution, can become impractical for arbitrarily complex problems requiring large or intricate neural network architectures. The modular feed forward neural network (MFFN) architecture decomposes a problem among a number of independent task specific neural networks, and is suggested here as a means of managing neuroevolution for complex problems. We present an algorithm for evolving MFFN architectures based on the NeuroEvolution of Augmenting Topologies (NEAT) algorithm. The algorithm proposed here, denoted MFF-NEAT, outlines an approach to automatically evolving, attributing fitness values and combining the task specific networks in a principled manner.