26.11.2012 Views

Technical Sessions – Monday July 11

Technical Sessions – Monday July 11

Technical Sessions – Monday July 11

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

203, 1200-781, Lisboa, Portugal, cmourao@iseg.utl.pt, Enrique<br />

Benavent, Angel Corberan, Luis Gouveia, Leonor S.Pinto<br />

The Profitable Mixed Capacitated Arc Routing Problem is a generalization of<br />

the Profitable Arc Tour Problem. Tasks may be either mandatory or optional.<br />

Each task has an associated demand, a profit, a deadheading cost and a traversal<br />

time. The objective is to find a set of tours that maximize the total net profit,<br />

respecting a maximal tour length and the vehicles capacity. The profit over a<br />

link is available only once, when the service is performed. We present compact<br />

flow based models for the PMCARP. Valid inequalities are also presented. The<br />

quality of the models is tested and analysed.<br />

3 - Optimization of Bus Schedules in Event Tourism using<br />

Mixed-integer Linear Programming<br />

Gianluca Brandinu, Department of Business Administration,<br />

University of Bern, Schuetzenmattstrasse 14, 3012, Bern,<br />

Switzerland, gianluca.brandinu@pqm.unibe.ch, Norbert<br />

Trautmann<br />

We study a bus tour of Bollywood-movie locations in Switzerland. The tour<br />

operator runs up to five busses per day; however, two or more busses cannot<br />

stay at the same location simultaneously. The planning problem consists in<br />

computing a feasible schedule for each bus such that a lexicographic combination<br />

of the total waiting and the total travel time is minimized. We formulate<br />

this problem as an MILP, and we report on numerical results obtained with the<br />

Gurobi Solver. We demonstrate how the computational effort can be reduced<br />

by introducing symmetry-breaking constraints.<br />

4 - Optimized Long-term Freight Infrastructure Planning<br />

Simon Dunstall, Mathematics, Informatics and Statistics,<br />

CSIRO, Private Bag 33, 3169, South Clayton, Victoria, Australia,<br />

Simon.Dunstall@csiro.au, Andreas Ernst, Kim Levy, Stuart<br />

Woodman, Olena Gavriliouk, Andrew Higgins, Martin<br />

Savelsbergh, Gaurav Singh<br />

The improvement of transport infrastructure for minerals freight continues in<br />

Australia, particularly in Queensland where the government and CSIRO have<br />

constructed a freight network optimization system. This system is known as<br />

the "Infrastructure Futures Analysis Platform", has a GIS interface and optimizes<br />

the capacity of roads, railways and facilities yearly over a 25 year horizon<br />

while satisfying freight demand and various practical constraints. We describe<br />

the mathematical model and our experience in working with GIS packages and<br />

freight transport policymakers.<br />

� MC-07<br />

<strong>Monday</strong>, 16:00-17:30<br />

Meeting Room 106<br />

OR Software<br />

Stream: OR software<br />

Invited session<br />

Chair: Pim Beers, Paragon Decision Technology, Julianastraat 30,<br />

2012ES, Haarlem, Netherlands, p.beers@aimms.com<br />

1 - Tackling Large-scale Optimization Problems within a<br />

Python-based Modeling Environment (Pyomo)<br />

William Hart, Data Analysis & Informatics, Sandia National<br />

Laboratories, P.O. Box 5800, 87185, Albuquerque, NM, United<br />

States, wehart@sandia.gov, John Siirola, Jean-Paul Watson<br />

The Pyomo modeling package can be used to formulate optimization models<br />

natively within the Python scripting language. Pyomo is an open-source modeling<br />

language that is being actively developed to support COIN-OR users in a<br />

variety of application areas. This talk will highlight recent developments in Pyomo<br />

and related Coopr packages, focusing on the overall speed and scalability<br />

of the modeling environment for large-scale optimization problems.<br />

2 - Attacking Hard Mixed-Integer Optimization Problems<br />

Using an Algebraic Modeling Language<br />

Robert Fourer, AMPL Optimization LLC, 2521 Asbury Avenue,<br />

60201-2308, Evanston, IL, United States, 4er@ampl.com<br />

IFORS 20<strong>11</strong> - Melbourne MC-08<br />

There are many tricks for formulating complex optimization models by use of<br />

integer variables, but what’s to be done when even the most advanced solvers<br />

can’t produce results in reasonable time? Substantial improvements in performance<br />

can often be achieved through carefully focused troubleshooting and experimentation.<br />

And sometimes, much better results can be achieved by "cheating"<br />

a bit on the formulation. A series of case studies illustrate that although a<br />

few general principles can offer guidance, much trial and error is involved, for<br />

which purpose a flexible modeling language is ideal.<br />

3 - AIMMS Modelling System - Demonstration<br />

Pim Beers, Paragon Decision Technology, Julianastraat 30,<br />

2012ES, Haarlem, Netherlands, p.beers@aimms.com<br />

AIMMS is an advanced modeling system for building optimization-based decision<br />

support applications. It is used by leading companies to support decision<br />

making and by universities for teaching operations research courses and for research<br />

projects. The talk will consist of a basic introduction of the modeling<br />

system AIMMS. We will give a global overview of the model representation,<br />

the graphical user interface, the solver and algorithmic capabilities. The presentation<br />

will contain a demonstration on how to build a small transport model<br />

in AIMMS.<br />

� MC-08<br />

<strong>Monday</strong>, 16:00-17:30<br />

Meeting Room 107<br />

Computational Methods in Biomolecular<br />

and Phylogenetic Analyses<br />

Stream: Bioinformatics<br />

Invited session<br />

Chair: Ming-Ying Leung, The University of Texas at El Paso, TX<br />

79968-0514, El Paso, United States, mleung@utep.edu<br />

1 - An Optimization Approach to Efficient Sampling over<br />

Phylogenetic Trees<br />

Russell Schwartz, Carnegie Mellon University, United States,<br />

russells@andrew.cmu.edu, Navodit Misra, Guy Blelloch, R Ravi<br />

Phylogenetics, or the inference of evolutionary trees, is commonly solved by<br />

Monte Carlo sampling over tree topologies, but prevailing methods depend on<br />

heuristic mixing time bounds that provide no guarantees of adequate mixing.<br />

We present a new strategy for generating provably well mixed samples efficiently<br />

in practice for important special cases of the problem. Our method<br />

reformulates the problem to create a rapidly mixing chain by performing a theoretically<br />

hard but practically efficient optimization on each sampling step.<br />

2 - Using Massively Parallel Ssequencing Data in Family<br />

Studies<br />

Melanie Bahlo, The Walter and Eliza Hall Institute of Medical<br />

Research, 3052, Parkville, VIC, Australia, bahlo@wehi.edu.au<br />

Massively parallel sequencing (MPS) data is making a strong impact in the<br />

field of disease gene identification, particularly in single gene disorders. We<br />

describe some of our recent work in extending methods using this new technology<br />

with two examples: (1) estimation of relatedness of individuals using<br />

RNA-seq data and (2) application of traditional statistical mapping techniques<br />

to MPS data to more effectively filter likely causal variants.<br />

3 - Comparative Studies in Biological Networks via Graphbased<br />

Features<br />

Kwok Pui Choi, National University of Singapore, <strong>11</strong>9077,<br />

Singapore, Singapore, stackp@nus.edu.sg<br />

Biomolecules orchestrate higher cellular functions. Recent breakthroughs in<br />

biotechnology have generated much information on biomolecular interactions<br />

in a genomic scale. Mathematical and computational approaches prove to be<br />

indispensible in providing deeper insights into these networks. Various graphbased<br />

measures, including the well known node degree distribution and clustering<br />

coefficient, have been introduced. We will explore the relationships among<br />

these measures and how they can be collectively applied to assess which evolutionary<br />

network model best describe a given biological network.<br />

4 - A Data Analysis System for iTRAQ<br />

Penghao Wang, School of Mathematics and Statistics, University<br />

of Sydney, 2006, Sydney, NSW, Australia,<br />

23

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!