Performance of various BFGS and DFP implementations with limited precision second order information. (2003)
Type of ContentDiscussion / Working Papers
PublisherUniversity of Canterbury
- Engineering: Reports 
This paper supports the claim that there is no discernible advantage in choosing factorised implementations (over non–factorised implementations) of BFGS methods when approximate Hessian information is available to full machine precision. However the results presented in this paper show that a factorisation strategy has clear advantages when approximate Hessian information is available only to limited precision. These results show that a conjugate directions factorisation outperforms the other methods considered in this paper (including Cholesky factorisation) for both BFGS and DFP formulae.
KeywordsQuasi-Newton methods; BFGS; DFP; numerical stability
ANZSRC Fields of Research01 - Mathematical Sciences
RightsAll Rights Reserved
Showing items related by title, author, creator and subject.
The performance of available methods for computing the polynomial coefficients of the quadratic function approximation is evaluated. By comparing the numerical results to those obtained by symbolic methods, for a variety of functions, the direct solution of the matrix equation and a variety of recursive algorithms are all shown to be numerically unstable. AMS classification Balakrishnan, Easwaran; McInnes, A.W. (University of Canterbury. Mathematics, 1991)The performance of available methods for computing the polynomial coefficients of the quadratic function approximation is evaluated. By comparing the numerical results to those obtained by symbolic methods, for a variety ...
Steinke, T. A. (University of Canterbury. Mathematics and Statistics, 2008)This report discusses shared memory parallel algorithms. It explains the benefits and difficulties of parallelizing algorithms by means of some examples. The main examples are parallel algorithms for calculating a Cholesky ...
Berger, Arno; Hill, Theodore P. (University of Canterbury. Department of Mathematics and Statistics, 2005)Conditions are given for a l map T to be a Newton map, that is, the map associated with a differentiable real-valued function via Newton's method. For finitely differentiable maps and functions, these conditions can ...