Gauss-Newton optimization technique is a method with less computation time and rapid convergence speed, but it requires better initial conditions. It is a tickler to lose its role or to bring about local-extreme-value problems. Under the circumstances, an improved damping form of Gauss-Newton optimization technique is presented to avoid these problems. By using the diagonal damping matrix, the damping role will be changed according to the relative modified amounts of the inversion parameters. The good effect is obtained when this approach is used in the inversion problem of the high frequency electromagnetic wave logging.