PT Unknown
AU Sergio Escalera
TI Coding and Decoding Design of ECOCs for Multi-class Pattern and Object Recognition A
PY 2008
AB Many real problems require multi-class decisions. In the Pattern Recognition field,many techniques have been proposed to deal with the binary problem. However,the extension of many 2-class classifiers to the multi-class case is a hard task. Inthis sense, Error-Correcting Output Codes (ECOC) demonstrated to be a powerfultool to combine any number of binary classifiers to model multi-class problems. Butthere are still many open issues about the capabilities of the ECOC framework. Inthis thesis, the two main stages of an ECOC design are analyzed: the coding andthe decoding steps. We present different problem-dependent designs. These designstake advantage of the knowledge of the problem domain to minimize the numberof classifiers, obtaining a high classification performance. On the other hand, weanalyze the ECOC codification in order to define new decoding rules that take fullbenefit from the information provided at the coding step. Moreover, as a successfulclassification requires a rich feature set, new feature detection/extraction techniquesare presented and evaluated on the new ECOC designs. The evaluation of the newmethodology is performed on different real and synthetic data sets: UCI MachineLearning Repository, handwriting symbols, traffic signs from a Mobile Mapping System, Intravascular Ultrasound images, Caltech Repository data set or Chaga’s diseasedata set. The results of this thesis show that significant performance improvementsare obtained on both traditional coding and decoding ECOC designs when the newcoding and decoding rules are taken into account.
ER