Matheon-Workshop 2004 on
Poster jpg pdf
Algebraic Complexity Theory
Algorithmic Differentiation

Thursday, December 9, pm. - Friday, December 10, pm.

Humboldt-Universität zu Berlin
Campus Adlershof
Rudower Chaussee 25
Room 1.410
Andreas Griewank
HU & Matheon
Bernd Bank
Trond Steihaug
Uni Bergen & HU & Matheon Guest

Institut für Mathematik
Math.-Nat. Fakultät II
Humboldt-Universität zu Berlin

DFG Research Center

Mathematics for key technologies

Hartwig Bosse
TU Berlin
Alin Bostan
Inria, Rocquencourt, France
Joos Heintz
Universidad de Buenos Aires
Shahadat Hossain
University of Lethbridge, Canada
Bernd Kummer
HU Berlin
Lutz Lehmann
HU Berlin
Uwe Naumann
RWTH Aachen
Eric Schost
Ecole polytechnique, Paris-Palaiseau
Jean Utke
Argonne National Laboratory, USA

It turns out that arithmetic circuits, i.e., directed acyclic graphs (DAGs) and straight line programs (SLPs) describing numerical computations, represent a basic data structure as well in automatic differentiation as in effective algebraic geometry. In this regard there exists a close connection between automatic differentiation, effective algebraic geometry and algebraic complexity theory, classic algebraic and Diophantine geometry, constraint data base theory and formal specification and transformation of programs.
The workshop aims at bringing together researchers in automatic differentiation, effective algebraic geometry, network optimization, and related fields in order to foster communication between the corresponding communities.
Timetable (tentative)

Thursday, December 9
13.00 - 14.00
A. Griewank
On whether, why and how algorithmic differentiation needs (more) algebraic complexity theory
14.00 - 15.00 J. Heintz The queries of algebraic geometry. Specification and data types in elimination theory.
15.00 - 16.00
Discussion, coffee break

16.00 - 16.30
U. Naumann
Complete and partial Jacobian accumulation by vertex, edge, or face elimination
16.30 - 17.00
J. Utke
DAG modifications, preserving optimalities and maintaining symmetries
17.00 - 17.30

17.30 - 18.00
E. Schost
A transposition theorem for RAM machines
18.00 - 18.30
A. Bostan
Improving the complexities of polynomial evaluation, interpolation and extrapolation: Tellegen's principle at work
18.30 - 19.30

19.30 -

Friday, December 10
09.00 - 09.30
B. Bank
Real elimination and arithmetic circuits
09.30 - 10.00
L. Lehmann
BKK-bounds for polar varieties
10.00 - 11.00
Discussion, coffee break

11.00 - 11.30
B. Kummer Newton methods for stationary points: regularity and auxiliary problems
11.30 - 12.00
H. Bosse
Representing polyhedra using few polynomial inequalities
12.00 - 13.00 Lunch break

13.00 - 13.30
S. Hossain
Exploiting derivative sparsity by matrix compression
13.30 - 14.00
T. Steihaug Higher order methods and the quest for order
14.00 - 15.00
Discussion and summary


For registration, please contact Ramona Klaass-Thiele until the end of November.