Effective Comparison of Unconstrained Optimization Techniques

The paper presents a means of attempting to account for overhead, as well as function evaluations, in evaluating unconstrained optimization techniques. Criteria are established which compare techniques on classes of algorithms. While not completely machine independent, and certainly not programmer independent, the new method eliminates much of the machine dependency of earlier criteria. The criteria are applied to three known and one new quasi-Newton algorithm, with interesting results.

• 저널Management Science · 22(3) · 321–330
• 토픽Numerical Methods and Algorithms · Computational Theory and Mathematics
• DOI10.1287/mnsc.22.3.321
• 저자David F. Shanno, K. H. Phua