Cambridge Systems Associates Limited (CSA), a small high end financial services analytics, consulting and software company (see www.cambridge-systems.com), has two 2 year postdoctoral positions, funded by the EU and starting as soon as possible.
The full time R&D work will involve development of CSA's iALM financial planning system and its patented STOCHASTICS stochastic optimization analyst software in high performance computing applications. Publications from the work will be expected.
Lloyds Banking Group and Citigroup are the financial partners and Numerical Applications Group (NAG), Maxeler and several leading European Universities are other partners on the EU High Performance Computing in Finance (HPC Finance) project. Please see the project website: www.hpcfinance.eu for a full list of partners, where CSA's detailed job descriptions may also be found (dates on these latter are irrelevant).
We are seeking recent PhDs (not more than 4 years from doctoral completion) in mathematics, statistics, OR, computer science or quantitative finance. Candidates must not be UK citizens, nor have spent more than one year in the UK to date.
The successful applicants will be CSA employees and salaries are generous, with moving allowances available. The active and exciting environment for these positions is Cambridge University. The company's elegant offices are in the centre of the University and the city.
Interested candidates should apply to Professor M.A.H. Dempster (mahd2@cam.ac.uk) at their earliest convenience.
Tuesday, September 18, 2012
Thursday, September 13, 2012
An Effective and Time-efficient Approach to Solving Linear Discrete Optimization Problems using Discretized Network Flow
A recent NSF grant is award to Shantanu Dutt. Here is the abstract below from the NSF website
Discrete optimization finds widespread use in almost all areas of human endeavor ranging from science to technology to business, and encompassing diverse applications such as chip design, power system design, robotics, bioinformatics, transportation, financial computing and industrial engineering. However, currently there is no general discrete optimization solver that can solve hard problems near-optimally in tractable runtimes (fast to moderate runtimes). To rectify this, this project will explore developing novel and efficient techniques for solving the class of 0/1 integer linear programming (ILP) problems that can be used to model a wide range of discrete optimization problems (DOPs). The approach used for this purpose is a solution technique termed discretized network flow (DNF), in which classical network flow (that solves a class of continuous linear programming problems), is constrained by special discrete requirements on the flow to yield valid solutions to 0/1 ILPs.
A successful completion of this project will yield algorithms and a software tool for solving large 0/1 ILP problems near-optimally and much faster than current techniques. This will represent a significant advance in the state-of-the-art of such solvers, and can be used to solve large DOPs more accurately and faster in many application areas ranging from genomics to chip design to robotics. This can help in answering fundamental issues in these application areas that have not been attempted so far, and also lead to better products and services in these areas. For example, in the area of chip design, the use of our solver can lead to much lower power and higher quality chips (e.g., with good performance and reliability) than are possible with current CAD tools, and thereby result in better and greener electronic products in several consumer and commercial application areas.
Discrete optimization finds widespread use in almost all areas of human endeavor ranging from science to technology to business, and encompassing diverse applications such as chip design, power system design, robotics, bioinformatics, transportation, financial computing and industrial engineering. However, currently there is no general discrete optimization solver that can solve hard problems near-optimally in tractable runtimes (fast to moderate runtimes). To rectify this, this project will explore developing novel and efficient techniques for solving the class of 0/1 integer linear programming (ILP) problems that can be used to model a wide range of discrete optimization problems (DOPs). The approach used for this purpose is a solution technique termed discretized network flow (DNF), in which classical network flow (that solves a class of continuous linear programming problems), is constrained by special discrete requirements on the flow to yield valid solutions to 0/1 ILPs.
A successful completion of this project will yield algorithms and a software tool for solving large 0/1 ILP problems near-optimally and much faster than current techniques. This will represent a significant advance in the state-of-the-art of such solvers, and can be used to solve large DOPs more accurately and faster in many application areas ranging from genomics to chip design to robotics. This can help in answering fundamental issues in these application areas that have not been attempted so far, and also lead to better products and services in these areas. For example, in the area of chip design, the use of our solver can lead to much lower power and higher quality chips (e.g., with good performance and reliability) than are possible with current CAD tools, and thereby result in better and greener electronic products in several consumer and commercial application areas.
Optimization without Round-off Errors
Recently, a NSF grant is awarded to Erick Moreno-Centeno on Optimization without Round-off Errors. It will be thrilling in the community since round-off erros have been a serious concern to solve large scale optimization problems, especially those with integer variables.
This Early-concept Grant for Exploratory Research (EAGER) provides funding to further develop an algorithm to solve systems of linear equations with a specific advantage over Gaussian Elimination in that the algorithm does not have round-off errors. The algorithm will be tailored and adapted to be used within existing optimization algorithms. The key property of the algorithm being further developed is that it maintains the integrality of the numbers throughout its execution. In particular, although the algorithm does contain divisions (which is crucial for the polynomial-time complexity of the
algorithm) all the divisions have a remainder of zero. A detailed complexity analysis of the algorithm will be performed; specifically, both its running time and the number of bits required to represent the numbers throughout the algorithm's execution will be investigated. A modification of the algorithm to perform LU factorizations will also be studied. Finally, the algorithm will be adapted to enable its use for the basis updates performed in state-of-the-art implementations of the revised simplex method.
If successful, the outcomes of this research will lead to a deeper understanding of computational linear algebra free of round-off errors and the associated applications within optimization algorithms. In the linear programming (LP) context, this research will enable us to solve large-scale LPs exactly; this, in turn, will significantly improve the accuracy and speed of integer programming solvers. In addition, since solving systems of linear equations is a critical subroutine of many numerical algorithms, this research will impact many application areas where linear systems frequently arise, including economics, physics, chemistry, and engineering.
This Early-concept Grant for Exploratory Research (EAGER) provides funding to further develop an algorithm to solve systems of linear equations with a specific advantage over Gaussian Elimination in that the algorithm does not have round-off errors. The algorithm will be tailored and adapted to be used within existing optimization algorithms. The key property of the algorithm being further developed is that it maintains the integrality of the numbers throughout its execution. In particular, although the algorithm does contain divisions (which is crucial for the polynomial-time complexity of the
algorithm) all the divisions have a remainder of zero. A detailed complexity analysis of the algorithm will be performed; specifically, both its running time and the number of bits required to represent the numbers throughout the algorithm's execution will be investigated. A modification of the algorithm to perform LU factorizations will also be studied. Finally, the algorithm will be adapted to enable its use for the basis updates performed in state-of-the-art implementations of the revised simplex method.
If successful, the outcomes of this research will lead to a deeper understanding of computational linear algebra free of round-off errors and the associated applications within optimization algorithms. In the linear programming (LP) context, this research will enable us to solve large-scale LPs exactly; this, in turn, will significantly improve the accuracy and speed of integer programming solvers. In addition, since solving systems of linear equations is a critical subroutine of many numerical algorithms, this research will impact many application areas where linear systems frequently arise, including economics, physics, chemistry, and engineering.
Tuesday, September 11, 2012
postdoctoral position in Carnegie Mellon University
A postdoctoral position is available in my group in the area of computational global optimization. I am looking for a person interested to work on the development and implementation of polyhedral cutting plane techniques for nonlinear and mixed-integer nonlinear problems in the context of the BARON project. For current group activities, see http://archimedes.cheme.cmu.ed
Best wishes, Nick Sahinidis
Best wishes, Nick Sahinidis
Subscribe to:
Posts (Atom)