The relevance vector machine technique for channel equalization application

S. Chen, S.R. Gunn, C.J. Harris
2001 IEEE Transactions on Neural Networks  
1529 [using (1)] and (n) = 0:0193 [using (7) ] were in good accordance wth the empirically measured value (5) (n) = 0:0157. Again, note that the high number of available patterns privileges the GF-based approximation. V. CONCLUSION This letter has extended Vapnik's theory by exploiting basic properties of KWMs. The validity of the involved research consists in a general procedure allowing one to estimate a classifier's VC-dim in the multiclass case: one need only repeat the procedure
more » ... in Section IV-B for the proper value of Nc and then apply expression (1). An additional, peculiar result of the presented approach lies in the opportunity given by (7) to estimate a classifier's GF. Abstract-The recently introduced relevance vector machine (RVM) technique is applied to communication channel equalization. It is demonstrated that the RVM equalizer can closely match the optimal performance of the Bayesian equalizer, with a much sparser kernel representation than that is achievable by the state-of-art support vector machine (SVM) technique.
doi:10.1109/72.963792 pmid:18249985 fatcat:atuxtsth3rgfvkr2rf3jeqblde