Output details

Multi-objective learning of Relevance Vector Machine classifiers with multi-resolution kernels

<24> This paper appears in a highly ranked journal with an impact factor of 3.3. It tackles the problem of learning classifiers using evolutionary multi-objective optimisation of the receiver operating characteristic. This widely used process is prone to overfitting and this paper investigates methods for effectively controlling the generalisation error. It is investigated in the context of the relevance vector machine which is an efficient sparse classifier when used with multi-resolution kernels, but highly susceptible to overfitting in this case.