On Testing Algorithms for Mathematical Programming Problems.
Abstract
This paper considers the comparative evaluation of algorithms for mathematical programming problems. It is concerned with the measurement of computational speed and examines critically the concept of equivalent number of function evaluations N sub e. Does this quantity constitute a fair way of comparing different algorithms. The answer to the above question depends strongly on whether or not analytical expressions for the components of the gradient and the elements of the Hessian matrix are available. It also depends on the relative importance of the computational effort associated with algorithmic operations vis-a-vis the computational effort associated with function evaluations. From the analyses performed and the results obtained, it is inferred that the direct measurement of the CPU time is still the more reliable way of comparing different minimization algorithms.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1976
- Accession Number
- ADA046252
Entities
People
- A. K. Wu
- Angelo Miele
- S. Gonzalez
Organizations
- Rice University