Support vector machine is new machine learning technique developped from the middle of 1990s.Being different from traditional neural network

it is based on structure risk minimization principle

while the latter on empirical risk minimization principle.A large number of experiments have shown that

comparing with traditional neural network

support vector machine has not only simpler structure

but also better performances

especially better generalization ability.In this paper

some new developments on support vector machine are introduced so as to draw our attention.