On what tasks does neuroevolution outperform basic application of neural networks or genetic algorithms?

Question

There has been a recent interest in combining genetic algorithms and neural networks into a general neuroevolution framework. The basic idea, is that your genetic algorithm is evolving the parameters of many neural-network which are then used to solve your task at hand. A sort of genetic programming but instead of evolving a snippet of code to do some task, you are evolving a neural network.

When should I use this combined approach instead of using neural-networks or genetic algorithms by themselves? For what types of problems has the combined approach produced better results that the individual approaches? For what types of problems is the combined approach the 'best' approach?

Answer 1

This has been researched for 20 years or so, and there are many papers claiming to outperform backpropagation. Xin Yao did a lot of work on this in the 1990s, and Kenneth Stanley created one of the currently most active frameworks, NEAT (NeuroEvolution of Augmenting Topologies (see http://www.cs.ucf.edu/~kstanley/neat.html and http://tech.groups.yahoo.com/group/neat/).

There's a lot of published material on different neuroevolutionary techniques, but these references may be useful in getting a feel for progress over the years:

1. Azzini, A., Tettamanzi, A. (2008) 'Evolving Neural Networks for Static Single-Position Automated Trading', Journal of Artificial Evolution and Applications, Volume 2008, Article ID 184286
2. Hintz, K.J., Spofford, J.J. (1990) 'Evolving a Neural Network', Proceedings, 5th IEEE International Symposium on Intelligent Control, pp. 479-484
3. Miller, G.F., Todd, P.M., Hedge, S.U. (1989) 'Designing neural networks using genetic algorithms', Proceedings of the Third International Conference on Genetic Algorithms
4. Montana, D.J. (1995) 'Neural Network Weight Selection Using Genetic Algorithms', Intelligent Hybrid Systems
5. Yao, X. (1993) 'Evolutionary artificial neural networks', International Journal of Neural Systems, Vol. 4, No. 3, pp. 203-222