TITLE:

High Performance Computational Differentiation Algorithms for Generalized Optimization Applications

DATE:

Friday, March 2nd, 2018

TIME:

3:30 PM

LOCATION:

GMCS-314

SPEAKER:

Dr. Ahmad Bani Younes, Department of Aerospace Engineering, San Diego State University.

ABSTRACT:

Computational differentiation has existed since the 1960’s as a scientific and engineering
methodology for automatically generating sensitivity partial derivative models for stability
assessments and optimization studies. Two solution strategies have been developed for
building sensitivity models. The first approach uses the analyst software, i.e. expressed
in Fortran, C, or other languages, as a template for writing a sensitivity software solution
for the problem. Unfortunately, very large partial derivative model result. In theory, the
derived first-order model can then be used as a template for the second and higher order
sensitivity models. A further complicating feature of this approach is that the resulting
software models are not readable by analysts. The second approach for generating higher-order
sensitivity models invokes the use of operator-overloading and user defined data structures
for enabling the compilier to derive and code the sensitivity solution. Some of these
developments are and demonstrated via various generalized optimization applications.

HOST:

Dr. Satchi Venkataraman

DOWNLOAD:

PDF File