Maximizing Vector Distances for Purpose of Searching—A Study of Differential Evolution Suitability
This paper explores suitability of using of differential evolution for maximizing of weighted distances of vectors in a set of vectors. Increase in vector distances simplifies searching for the best matching vector what is a common task in many areas (for instance in biometric identification of people). Maximizing of weighted distances itself is complex and nonlinear problem. The differential evolution is efficient enough and helps in decreasing of the computational complexity space compared to enumerative methods where all possible combinations are calculated. To find out, if differential evolution can help with the problem, model experiments were introduced and executed. Experiments showed that differential evolution is able to resolve the problem.
Keywordsoptimization differential evolution distance metric nonlinear
Unable to display preview. Download preview PDF.
- 1.Jain, A.K., Ross, A.: Handbook of Biometrics. Springer (2008) ISBN 978-0-387-71040-2Google Scholar
- 2.Spall, J.C.: Introduction to Stochastic Search and Optimization. Wiley (2003) ISBN 0-471-33052-3Google Scholar
- 3.Price, K., Storn, R.M., Lampinen, J.A.: Differential evolution: A Practical Approach To Global Optimization. Natural computing series (1) (2005)Google Scholar
- 4.Deza, M.M., Deza, E.: Encyclopedia of Distances. Springer (2009) ISBN 3-642-00233-5Google Scholar
- 5.Jolliffe, I.T.: Principal Component Analysis, 2nd edn. Springer (2002) ISBN 0-387-95442-2Google Scholar
- 6.Storn, R.: Differential Evolution (DE), http://www1.icsi.berkeley.edu/~storn/code.html (cited March 02, 2014)
- 7.Accord.Net Framework, http://accord-framework.net/ (got January 21, 2014)
- 8.Souza, C.: Principal Component Analysis in C#, http://crsouza.blogspot.cz/2009/09/principal-component-analysis-in-c.html (got January 21, 2014)