On the design of an ECOC-Compliant Genetic Algorithm

作者：

Highlights：

• A novel Genetic Algorithm to optimize the ECOC coding step is presented.

• The crossover and mutation operators are redefined taking into account the ECOC properties.

• A new operator that is able to extend the ECOC code is developed.

• We introduce a novel regularization parameter that is able to control the number of dichotomies.

摘要

Highlights•A novel Genetic Algorithm to optimize the ECOC coding step is presented.•The crossover and mutation operators are redefined taking into account the ECOC properties.•A new operator that is able to extend the ECOC code is developed.•We introduce a novel regularization parameter that is able to control the number of dichotomies.

论文关键词：ECOC,Genetic Algorithms,Multi-class classification

论文评审过程：Received 13 September 2012, Revised 6 June 2013, Accepted 16 June 2013, Available online 26 June 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.06.019