The global convergence of partitioned BFGS on problems with convex decompositions and Lipschitzian gradients
The main pupose of this paper is the extension of Powell's global convergence result to the partitioned BFGS method introduced by Griewank and Toint. Even in the unpartitioned case the original result is strengthened because the search directions need not be computed exactly and the gradient is only required to be Lipschitzian rather than differentiable. Using the PSI - functional of Byrd an Nocedal a strong form of R-superlinear convergence is obtained if the gradients are uniformly convex and strictly differentiable at the minimizer /sub *x/. In order to deal with the possibility of singular functions we utilize a damping of the BFGS update that becomes inactive if the problem turns out to be regular near x/sub *x/. 30 refs.
- Research Organization:
- Argonne National Lab., IL (USA)
- DOE Contract Number:
- W-31109-ENG-38
- OSTI ID:
- 5434080
- Report Number(s):
- ANL/MCS-TM-105; ON: DE88006249
- Country of Publication:
- United States
- Language:
- English
Similar Records
On line search termination criteria for and convergence of conic algorithms for minimization
A tool for the analysis of quasi-Newton methods with application to unconstrained minimization