The Computer Can Differentiate Too
TEXAS UNIV AT AUSTIN CENTER FOR CYBERNETIC STUDIES
Pagination or Media Count:
For mathematical problems whose solution requires the evaluation of first or second partial derivatives of user-defined functions, symbolic differentiation is offered as a viable alternative to numerical differencing techniques and user-written subprograms. The paper describes an efficient user- oriented approach, equivalent to symbolic differentiation, for the generation of code capable of evaluating partial derivatives of user-specified functions. Computational results of an experimental implementation of the method are reported. The feasibility of this approach is additionally demonstrated by interfacing the experimental code with the Sequential Unconstrained Minimization Technique of Fiacco and McCormick for the solution of nonlinear programming problems and results are reported.
- Theoretical Mathematics
- Computer Programming and Software
- Computer Hardware