GENETIC PROGRAMMING FOR MULTI-LABEL CLASSIFICATION
BASIC INFORMATION
Ph.D. Student: José Luis Ávila
Advisors: Eva L. Gibaja, Sebastián Ventura
Defended on: June 2013
Keywords: multi-label classification, gene expression programming, discriminant functions, rules
Digital version
DESCRIPTION
The main objective of this thesis is the development of a series of genetic programming models to solve multilabel classification problems. To do it, we propose the development of a multi-label classification using a genetic programming approach that allows working with numerical, categorical and nominal data sets.
The classification problem is to associate a series of labels to a number of examples or patterns. In the traditional classification of each training pattern can only associate a single label.
In multi-label classification, the objectives are not disjoint sets with each other, and may have patterns that are associated with more than one label. Therefore, the examples are associated with a set of labels Y i L, and the approximate function f (x) may take various values within the set of labels.
On the other hand, evolutionary algorithms are search algorithms that maintain a set or population of candidate solutions or individuals, which are applied a series of genetic operators, and selecting at each iteration the best individuals to generate the population the next generation.
Among the highlights evolutionary algorithms, genetic programming paradigm has been successfully used in traditional classification problems, where each individual represents a classifier or procedure for classifying a set of patterns.
However, genetic programming has not been used to solve multi-label classification, which is the starting point of work targeted.
FUNDS
The development of this thesis was supported by:
- Regional Government of Andalusia, project P08-TIC-3720.
- Inter-minisiterial Commision for Science and Technology (CICYT), project TIN2008-06681-C06-03.
PUBLICATIONS ASSOCIATED WITH THIS THESIS
INTERNATIONAL JOURNALS
- J.L. Ávila, E.L. Gibaja, A. Zafra and S. Ventura. A Gene Expression Programming Algorithm for Multi-Label Classification.Journal of Multiple-Valued Logic and Soft Computing, vol. 17(2-3), pp. 183-206. 2011.
INTERNATIONAL CONFERENCES
- J.L. Ávila, E.L. Gibaja, A. Zafra and S. Ventura. A niching algorithm to learn discriminant functions with multi-label patterns.International Conference on Intelligent Data Engineering and Automated Learning, pp. 570-577. 2009.
- J.L. Ávila, E.L. Gibaja, A. Zafra and S. Ventura. Multi-label classification with gene expression programming. International Conference on Hybrid Artificial Intelligence Systems, pp. 629-637. 2009.
- J.L. Ávila, E.L. Gibaja and S. Ventura. Evolving multi-label classification rules with gene expression programming: a preliminary study International Conference on Hybrid Artificial Intelligence Systems, pp. 9-16. 2010.