We empirically evaluate several state-of-theart methods for constructing ensembles of heterogeneous classifiers with stacking and show that they perform (at best) comparably to se...
Researchers in commonsense, qualitative spatial and temporal reasoning (QSTR) provide flexible and intuitive methods for reasoning about vague and uncertain information including ...
The kernel-parameter is one of the few tunable parameters in Support Vector machines, controlling the complexity of the resulting hypothesis. Its choice amounts to model selection...
Nello Cristianini, Colin Campbell, John Shawe-Tayl...
Based on independent component analysis (ICA) and self-organizing maps (SOM), this paper proposes an ISOM-DH model for the incomplete data’s handling in data mining. Under these ...
Abstract--A crucial issue in designing learning machines is to select the correct model parameters. When the number of available samples is small, theoretical sample-based generali...