Quantitative evaluation of dependence among outputs in ECOC classifiers using mutual information based measures
Francesco Masulli, Giorgio Valentini · 2002
In previous work by the authors (2000), it has been experimentally shown that the implementation of error correcting output coding (ECOC) classification methods with an ensemble of parallel and independent nonlinear dichotomizers (ECOC PND) outperforms the implementation with a single monolithic multilayer perceptron (ECOC MLP). The low dependence of the errors on different codeword bits was qualitatively indicated as one of the main factors affecting this result. In this paper, they quantitatively evaluate the dependence of output errors in ECOC learning machines using mutual information based measures, and we study the relation between dependence of output errors and classification performances.