Particle Swarm Optimization with Interval-valued Genotypes and Its Application to Neuroevolution

The author proposes an extension of particle swarm optimization (PSO) for solving interval-valued optimization problems and applies the extended PSO to evolutionary training of neural networks (NNs) with interval weights. In the proposed PSO, values in the genotypes are not real numbers but intervals. Experimental results show that interval-valued NNs trained by the proposed method could well approximate hidden target functions despite the fact that no training data was explicitly provided.

• Hidehiko Okada

• Evolutionary algorithms
• swarm intelligence
• particle swarm optimization
• neural network
• interval arithmetic.

[1] H. Ishibuchi, H. Tanaka and H. Okada, An architecture of neural
networks with interval weights and its application to fuzzy regression
analysis, Fuzzy Sets and Systems, vol.57, no.1, pp.27-39, 1993.
[2] D.B. Fogel, L.J. Fogel and V.W. Porto, Evolving neural networks,
Biological Cybernetics, vol.63, issue 6, pp.487-493, 1990.
[3] X. Yao, Evolving artificial neural networks, Proc. of the IEEE, vol.87,
issue 9, pp.1423-1447, 1999.
[4] K.O. Stanley and R. Miikkulainen, Evolving neural networks through
augmenting topologies, Evolutionary Computation, vol.10, no.2,
pp.99-127, 2002.
[5] D. Floreano, P. Durr and C. Mattiussi, Neuroevolution: from architectures
to learning, Evolutionary Intelligence, vol.1, no.1, pp.47-62, 2008.
[6] H. Okada, Proposal of fuzzy evolutionary algorithms for fuzzy-valued
genotypes, Proc. of International Conference on Instrumentation, Control, Information Technology and System Integration (SICE Annual
Conference) 2012, pp.1538-1541, 2012.
[7] G. Alefeld and J. Herzberger, Introduction to Interval Computation,
Academic Press, 1983.