Hybrid Spectral Conjugate Gradient Method Based on Strong Wolfe Line Search
The conjugate gradient method has the advantages of small storage and fast computation. This article designs a new hybrid conjugate parameter with convex combinations based on the conjugate parameter types of PRP classes. This parameter not only has the good convergence properties of FR method, but also has good numerical results of PRP method. At the same time, corresponding spectral conjugation parameters are designed based on the new parameters. This article proves that the algorithm has global convergence with strong Wolfe line search conditions. Finally, through numerical experiments on the problems in the CUTEr test set, it is found that the algorithm has good numerical performance.
万方数据
维普
