A Trust Region Algorithm With Reduced Dimension Of Trial Step For Equality Constrained Optimization

A trust-region algorithm for equality constrained optimization problem is presented, This algorithm uses the Byrd and Omojokun way of computing the trial steps, and decompose the trial steps into two components: normal components and tangential components. But it differs from the Byrd and Omojokun algorithm in using a falling dimension way in computing the tangential components which similar to the J.Pbulteau and J.Pvial method in computing unconstrained optimization problems. This method needn the condition of linear independence on the gradients of the constraints. Global convergence is proved under some supposed conditions. Some examples is presented in the paper.