Large Scale Nonlinear and Semidefinite Programming

• The workshop will
  follow the CORS/INFORMS Joint International Meeting May 16-19, at The Banff Centre, Banff, Alberta. And, it will
  precede the IPCO at New York University June 9-11. and the CMS and CAIMS joint conference in Halifax, N.S. June 12-15.
  We hope that our conference provides incentive for participants to join the other three conferences and conversely.
• The workshop will follow the same format as the highly successful Workshop on: Novel Approaches to Hard Discrete Optimization, held April 2001. We plan on having 10 invited speakers with approximately 20 contributed talks (no parallel sessions). We expect 40-50 external participants and 20-30 local participants.
  There will be a special session for graduate students to present their work. (We are currently discussing the possiblity of a short course as well.)
• The aim or scope of the workshop is to bring together researchers from several communities such as:
  • algorithmic nonlinear optimization
  • combinatorial optimization, dealing with computational methods for NP-hard problems
  • computer scientists interested in scientific parallel computing,
  who share a common interest to do computations on (large-scale) hard combinatorial optimization problems.
  Topics to be covered include:
  • General Nonlinear Programming Problems
  • Semidefinite programming and reformulation schemes
  • Quadratic and multi-dimensional assignment problems
  • Euclidean Distance Matrix Problems (Molecular Conformation)
  • Boolean quadratic programming
  • Massive graph problems
  • Massively parallel distributed processing
  • VLSI design

Theme/Rationale

A rich source of large-scale nonlinear problems arises from relaxations of NP-hard combinatorial optimization problems. The success to solve Semidefinite Programs (of moderate size) has resulted in increased interest in higher order relaxations of combinatorial optimization problems, which give tighter relaxations but which are computationally challenging.

Recent progress towards large-scale nonlinear programming is mostly based on algorithmic ideas which avoid using standard interior-point technology. These include first order methods, algorithms which focus on the sparsity structure as well as nonsmooth methods applied to some Lagrangian dual. In addition, several parallel implementations of IPM's as well as column generation-type techniques have been introduced and show promise.

It is the purpose of this workshop to provide a forum to exchange ideas on large sparse programming between researchers in general Nonlinear Programming and those in Cone Programming.

The workshop aims in particular at researchers from the following communities

• algorithmic nonlinear optimization
• combinatorial optimization, dealing with computational methods for NP-hard problems
• computer scientists interested in scientific parallel computing, who share a common interest to do computations on (large-scale) hard combinatorial optimization problems.

Budget (other funding sources)

There are currently no other sources of funding.
We include a budget request for travel for foreign speakers. However, we realize
that this may not be realistic and that we will have to ask speakers to
pay their own travel costs. However, we would like to help with
accomodations and food.

The estimates are for  30 speakers while total participants number 60.0 
------------------------------------- 
Foreign speakers travel costs (14-16): $14000.00
Student participants travel costs (14-16): $4000.00 
Accomodations for speakers for 4 nights at 46.20 per plus taxes: $5932.08 
Reception at $5.65 per  $339.00 
Coffee at $2.55 per  $459.00 
Breakfast at $9.00 per $1620.00 
Lunch at $12.50 per $2250.00 
Banquet at $20.95 per plus $65 setup cost $1322.00 
incidentals (publishing, tags, etc..): $1030.00 
streaming video record: $1000.00 
------------------------------------- 
Total costs $31952.08