2nd Latin American
School and Workshop on Polynomial Systems
Angra dos Reis, RJ, Brasil - Feb 28 to March 4, 2005
Description and scope
The theoretical
study of the solution set of polynomial equations gave rise to classical
areas in mathematics, such as Algebra or Algebraic Geometry. Polynomial systems
arising from practical or industrial applications require a much finer study
of the system structure than what is available from classical tools.
This conference is interdisciplinary
due to the nature of its object. The analysis of equation structure involves,
besides Algebraic Geometry, concepts of Convex Geometry, Symplectic Geometry,
Complexity, Numerical Analysis, etc.. In addition, efficient algorithms to
solve systems of polynomial equations combine tools and concepts of
Computational Algebra, Computational Algebraic Geometry, Differential Geometry,
Linear Programming, Sparse Numerical Linear Algebra, Combinatorics, etc..
Applications of polynomial systems arise in many areas,
such as mechanical engineering, chemistry and bio-chemistry, control theory,
computer graphics, mathematical physics. Special attention here will be given
to applications to statistics.
Target audience
The school is
intended for a public ranging from first year graduate students in mathematics
to research scientists. There will be a short course and five survey lectures
during the morning.
The scientific program is intended
for specialists in the subject, or researchers and students wishing
to enter the subject.
Five lectures by Bernd Sturmfels
Algebraic Geometry of Statistical Models
The five lectures offer an introduction
to algebraic methods in statistics. Many statistical models for discrete
data, in particular those used in machine learning and computational biology,
can be regarded as algebraic varieties inside a high-dimensional probability
simplex. While some of these models correspond to classical geometric objects
(e.g. Segre varieties and their secant varieties), others pose new and unexpected
challenges for algebraic geometers. Conversely, computational
algebraic geometry has a lot to offer for statistical analysis since certain
key problems, such as computing maximum likelihood estimates and checking
identifiability, lead to systems of polynomial equations.
Tentative Outline:
Lecture 1 Graphical Models
Lecture 2 Maximum Likelihood Equations
Lecture 3 Mixture Models and Secant Varieties
Lecture 4 The Expectation Maximization Algorithm
Lecture 5 Tropicalization and Polytope Propagation
Survey lectures
Alicia Dickenstein: Toric Ideals and Applications
Gregorio Malajovich: Mixed volume and sparse homotopy
Laurent Busé: Elimination Theory, Commutative Algebra
and Applications
Ioannis Emiris: Sparse Elimination and Geometric Applications
Jean-Pierre Dedieu: Newton iteration and applications.
There will also be invited (to be announced soon) and
contributed talks. The deadline for submissions is October 31, 2004.
Financial support for participants
Participants are supposed to cover
their own expenses. We are working on obtaining support for students and
researchers from Brasil and from the Mercosul community.
Local information.
The meeting will be held at the Portogalo
Suíte Hotel, Estrada Rio Santos km 71, Angra dos Reis, RJ, Brasil.
(Distance to Rio de Janeiro: 130 km. Distance to Angra dos Reis: 15km).
Previous conference.
The first Latin American School and
Workshop on Polynomial Systems was held in Buenos Aires, from 14 to 26 of
july 2003, with approximately 70 registered participants.
Organizing Committee
Alicia Dickenstein
(co-chair), Departamento de Matemática, FCEyN, Universidad de
Buenos Aires.
Alvaro Rittatore,
Centro de Matemática, Universidad de la República, Uruguay.
Gregorio Malajovich
(co-chair), Departamento de Matemática Aplicada, Universidade
Federal do Rio de Janeiro.
Ioannis Emiris,
Department of Informatics & Telecommunications, National University
of Athens.
Vilmar Trevisan,
Instituto de Matemática, Universidade Federal do Grande do Sul.
Homepage : http://www.labma.ufrj.br/~gregorio/conferences/polysys
Contact e-mail: polysys@labma.ufrj.br