Based on analyzing the quadratic programming of SVM and using the method of feasible directions to solve the quadratic programming,this paper solves the quadratic programming by the method of feasible directions and raises a improved SVM algorithm for unlinear.

At the same time,the traditional feasible direction method(FDM) was improved.

In this paper, a feasible direction method is used to resolve problems of which the storage space is limited, the least quantity of purchase is no less than the starting point and different goods must be placed separately.

Furthermore,the results of comparative calculation show that the new method can considerably improve the rate of convergence for the feasible direction method.

This paper introduces the approaching feasible direction algorithm (AFDA) for solving linearly constrained nonlinear programming problems.

Based on the semidefinite programming relaxation of the max-cut problem,a feasible direction algorithm is proposed,and its convergence is proved.

In this paper, the possilbe invalidation of zoutendijk method of feasible directions was analysed base on vector and its solution was given.

The Topkis-Veinott feasible direction method is applied for the solution of this optimisation and a new algorithm is proposed.

A strongly subfeasible directions method for solving nonlinear inequality constrained optimization problems is presented by using the generalized gradient projection.

Nonlinear equality and inequality constrained optimization is discussed,a new algorithm with arbitrary initial is presented by using the generalized projection technique and the idea of strongly subfeasible directions method This algorithm possesses global convergence and the search directions are strongly subfeasible descent directions Furthermore,better strong convergence is arrive