Global Convergence of the Partitioned Broyden’s Algorithm for Convex Partially Sepearable Optimizations

摘要

In this paper, the so—called partitioned Broyden’s algorithms used for solving Partially seperable optimization with a convex decomposition is concerned. Global convergence is proved for this type of "partitioned updating" quasi-Newton method. The algorithm is well adapted to unconstrained problems involving many variables.