Output details

Cohort-based kernel visualisation with scatter matrices

<24> Machine learning approaches. This originality of the paper is to extend to non-linearly separable data a previous methodology for cluster-based visualisation with scatter matrices (Lisboa et al ‘Cluster-based visualisation with scatter matrices’ Pattern Rec Letters 2008). The theoretical basis for non-linear modelling is the rigorous application of kernel methods. The significance of the paper is to enable efficient visualisation to be applied to class-labelled data, whereas the previous methodology was restricted to linearly separable data such as cluster partitions.