In the above description, the site meant that they used SMO to solve the quadratic programming (QP) problem that arose during training for vector support machines, since previously available SVM training methods were much more complicated and required expensive third-party QPs.
You can get a more detailed description of SMO here
source
share