Neural Computation
The nature of statistical learning theory
The nature of statistical learning theory
Moderating the outputs of support vector machine classifiers
IEEE Transactions on Neural Networks
The evidence framework applied to support vector machines
IEEE Transactions on Neural Networks
Soft-Sensor Method Based on Least Square Support Vector Machines Within Bayesian Evidence Framework
ISNN '07 Proceedings of the 4th international symposium on Neural Networks: Advances in Neural Networks, Part III
Hi-index | 0.00 |
Following previous successes on applying the Bayesian evidence framework to support vector classifiers and the Ɛ-support vector regression algorithm, in this paper we extend the evidence framework also to the ν-support vector regression (ν-SVR) algorithm. We show that ν-SVR training implies a prior on the size of the Ɛ-tube that is dependent on the number of training patterns. Besides, this prior has properties that are in line with the error-regulating behavior of ν. Under the evidence framework, standard ν-SVR training can then be regarded as performing level one inference, while levels two and three allow automatic adjustments of the regularization and kernel parameters respectively, without the need of a validation set. Furthermore, this Bayesian extension allows computation of the prediction intervals, taking uncertainties of both the weight parameter and the Ɛ-tube width into account. Performance of this method is illustrated on both synthetic and real-world data sets.