For example it's intended to optimize configurations of a ANN, or a Decision Tree, or any Supervised or Unsupervised model via a GA, PSO or any bioinspired technique. How can I choose well parameters like population, number of generations etc., for a GA, or other important parameters for a bioinspired techinique that's being used in order to perform well over that learning system?
Dear Inacio,
You can try the experimental design approach. Following article could be useful.
https://www.researchgate.net/publication/2559922_Tuning_PSO_Parameters_Through_Sensitivity_Analysis
Technical Report Tuning PSO parameters through sensitivity analysis
You can try automated algorithm configuration, it has been successfully applied to machine learning and the software is available:
http://www.cs.ubc.ca/labs/beta/Projects/SMAC/
Frank Hutter, Holger Hoos, and Kevin Leyton-Brown.
Bayesian Optimization With Censored Response Data
2011 NIPS workshop on Bayesian Optimization, Experimental Design, and Bandits.
Thank you for the answer, Peter. Quite interesting project, I'm gonna test it.
Dear Inacio,
You can try the experimental design approach. Following article could be useful.
https://www.researchgate.net/publication/2559922_Tuning_PSO_Parameters_Through_Sensitivity_Analysis
Technical Report Tuning PSO parameters through sensitivity analysis
Surely, Behrouz, really a quite interesting idea. Thank you for suggestion.
In my opinion, firstly you should look at papers which have done similar research and see what parameters they are suggesting, then you will have an idea and maybe try some experiments. Secondly, you should look at papers which deal with automatic algorithm configuration, including those one Shüller and Nedushan suggested. In this second step you will find options such as Fuzzy, Neural Networks, Meta-Evolutionary algorithms and so on. Then, you can pick up one, or even more than one, and try it in your algorithms.
This paper may help you.
https://www.researchgate.net/publication/263531240_Parameter_selection_in_particle_swarm_optimisation_A_survey?ev=prf_pub
Article Parameter selection in particle swarm optimisation: A survey
Parameter selection in GA
Boyabatli, O., & Sabuncuoglu, I. (2004). Parameter selection in genetic algorithms. Journal of Systemics, Cybernetics and Informatics, 4(2), 78.
Gates Jr, G. H., Merkle, L. D., Lamont, G. B., & Pachter, R. (1995, November). Simple genetic algorithm parameter selection for protein structure prediction. In Evolutionary Computation, 1995., IEEE International Conference on (Vol. 2, pp. 620-624). IEEE.
Kaiser Jr, C. E., Lamont, G. B., Merkle, L. D., Gates Jr, G. H., & Pachter, R. (1997, April). Exogenous parameter selection in a real-valued genetic algorithm. In Evolutionary Computation, 1997., IEEE International Conference on (pp. 569-574). IEEE.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.111.8696&rep=rep1&type=pdf
http://ink.library.smu.edu.sg/cgi/viewcontent.cgi?article=1840&context=lkcsb_research
Conference Paper Exogenous parameter selection in a real-valued genetic algorithm
Parameter selection in PSO
Zhang, W., Ma, D., Wei, J. J., & Liang, H. F. (2014). A parameter selection strategy for particle swarm optimization based on particle positions. Expert Systems with Applications, 41(7), 3576-3584.
Xu, M., & Gu, J. (2015, November). Parameter selection for particle swarm optimization based on Stochastic multi-objective optimization. In Chinese Automation Congress (CAC), 2015 (pp. 2074-2079). IEEE.
Sun, J., Fang, W., Wu, X., Palade, V., & Xu, W. (2012). Quantum-behaved particle swarm optimization: Analysis of individual particle behavior and parameter selection. Evolutionary computation, 20(3), 349-393.
Hasanzadeh, M., Meybodi, M. R., & Ebadzadeh, M. M. (2013). Adaptive Parameter Selection in Comprehensive Learning Particle Swarm Optimizer. In Artificial Intelligence and Signal Processing (pp. 267-276). Springer International Publishing.
http://www.cs.ox.ac.uk/people/vasile.palade/papers/Evol_Comp_Published_Sun_et_al.pdf
http://ceit.aut.ac.ir/~meybodi/paper/Hassanzadeh-Meybodi-Ebadzadeh-AISP-2013-FULL.pdf
- Badges
- Science topic