Nonlinear classifier combination for simple combination types
Mehmet Umut Şen, Hakan Erdoğan · 2011
Classifier combination has been an important research area because of their contribution to the accuracy and robustness. Supervised linear combiner types are shown to be strong combiners; but nonlinear types are not well investigated. In this work, we show a method to obtain non-linear versions of simple linear combiner types. Experiments are conducted on four different databases and results are examined. It is observed that we can obtain better accuracies with non-linear combinations for certain types.