I read the article "Successive Overrelaxation for Support Vector Machines" IEEE Transactions on Neural Networks, 10, 1999, 1032-1037. and I have tried to implement it. I have already an implementation for SMO and I would like to do another for SOR but my SOR convergence rate is very very slow when check against SMO. I did all steps cited in article above but I think I am doing something wrong. Do anybody have any pseudo-code, program or algorithm to help me in this task ? Thanks a lot Marcelo Barros http://www.litc.cpdee.ufmg.br/~barros/