Automatic Differentiation and Automatic Code Generation Techniques for Scientific Computing
Paul I. Barton, John E. Tolsma
Mon-Fri, Jan 7-11, 14-18, 10-11:00am, 66-360, 66-064 from 1-3pm each day-lab
Enrollment limited: first come, first served
Limited to 20 participants.
Participants requested to attend all sessions (non-series)
Prereq: See description
An Automatic (or Algorithmic) Differentiation (AD) tool takes a user's model coded in an imperative programming language, and from this automatically generates a new subroutine that will evaluate analytic partial derivative values for this model. The course is intended to be of interest to anyone at the Institute interested in computational science and engineering, and what AD tools can do for them. The theory and implementation of derivative value computation using AD will be covered, and extensions to other applications such as sparsity patterns, discontinuity handling, interval extensions, and convex relaxations. Lectures will be accompanied by electronic classroom sessions where students will get a chance to use the AD tool DAEPACK.
Web: http://yoric.mit.edu/AD
Contact: Paul I. Barton, 66-464, 253-6526, pib@mit.edu
Sponsor: Chemical Engineering
Latest update: 29-Oct-2001
|
|