Octave to MOSEK Interface

Octave is a scripting language and interactive environment suited for numerical computations. As part of our efforts to bring powerful optimization tools closer to our customers, MOSEK is pleased to announce an open source Octave to MOSEK interface is now available at Google Code Project Hosting. The interface makes it possible to solve

• linear,
• conic quadratic (aka. second order cone), 
• and mixed-integer 

problems from Octave. As any convex quadratic problem can be reformulated as a conic quadratic problem, the interface can also be employed to solve convex quadratic and quadratically constrained problems. Note that an R to MOSEK interface was released earlier this year.

MOSEK version 6.0.0.126

MOSEK version 6.0.0.126 is available for download at the MOSEK website. The major changes are:

• Fixed a bug in the primal simplex optimizer.
• Fixed a bug in the dual simplex optimizer.
• Fixed a bug in the presolve.
• Fixed a recently introduced bug in the network detection procedure.

MOSEK version 6.0.0.124

MOSEK version 6.0.0.124 is available for download at the MOSEK website. The major changes are:

• Upgrade FlexLM license license software to version 11.10.0.
• Fixed a bug in the network detection procedure.
• Fixed a bug in the convexity checker.
• Fixed performance issue in the mixed integer optimizer.
• Improved the interior point optimizers for linear problems for certain hard problems.

MOSEK version 6.0.0.122

MOSEK version 6.0.0.122 is available for download at the MOSEK website. The major changes are

• Several bug fixes to the mixed integer optimizer.
• Support for the MinGW tool chain.
• Support of 64-bit Python at the 64-bit MAC OSX platform.

R to MOSEK interface

R is popular tool for statistical computing that has many users in financial industry for example. Therefore, MOSEK is pleased to announce that an open source R to MOSEK interface is now available at R-forge. The interface makes it possible to use the powerful MOSEK optimizers for

• linear,
• conic quadratic (aka. second order cone), 
• and  mixed-integer 

problems from R. Note any convex quadratic problem can reformulated as a conic quadratic problem and hence the interface can be employed to solve convex quadratic and quadratically constrained problems.

New technical report about infeasible linear problems.

A new technical report about linear infeasible problems has been made available at the MOSEK website. The report is useful to  MOSEK users who wants to know more about the MOSEK infeasibility reports.

New website design

The MOSEK website has been redesigned so the layout is simpler. Also the search facility has been improved significantly. We hope you like it!

MOSEK version 6.0.0.119 is available for download.

MOSEK version 6.0.0.119 is available for download. The major changes are: 

• Fixed a bug in the dual simplex optimizer.
• Fixed several bugs in the mixed integer optimizer. 
• Fixed an issue in the interior-point optimizer for linear problems. 

MOSEK version 6.0.0.114 is available for download.

MOSEK version 6.0.0.114 is available for download. The major changes are: 

• Fixed a bug in the concurrent optimizer.
• Fixed several bugs in the mixed-integer optimizer.
• Removed debug output occuring when reading MOSEK binary task files.

Updated MOSEK connector to Microsoft Solver Foundation

The MOSEK connector to the Microsoft Solver Foundation has been updated to support MOSEK version 6. See the Solver Foundation blog for details.

MOSEK solves a large transportation problem

Erwin Kalvehagen of Amsterdam Optimization has an interesting blog post that demonstrates MOSEK can solve a huge transportation problem extremely fast. In fact GAMS almost the same time on setting up the problem as  MOSEK spends on solving it.

MOSEK version 6.0.0.106 is available for download.

MOSEK version 6.0.0.106 is available for download. The major changes are: 

• Fixed a bug in the conic optimizer that in special cases made MOSEK report an nonoptimal solution.
• Fixed several bugs in the mixed integer optimizer.
• Downgraded Flexlm to 11.9.0 on the UNIX platforms.
• Fixed a couple of bugs in the graph partitioning based ordering algorithm (employed in the interior-point optimizers). 
• Improved the network structure detection capabilities. 
• Fixed a bug in the basis identification procedure. This bug occurs very rarely.

MOSEK version 6.0.0.105 is available for download.

MOSEK version 6.0.0.105 is available for download. The major changes are:

• Fixed a couple of bugs in the graph partitioning based ordering algorithm (employed in the interior-point optimizers).
• Improved the network detection structure capabilities.
• Fixed a bug in the basis identification procedure. This bug occurs very rarely.

MOSEK version 6.0.0.103 is available for download.

MOSEK version 6.0.0.103 is available for download. The major changes are:

• Fixed several bugs in the mixed integer optimizer.
• Fixed a bug in the graph partioning based ordering that in rare cases could cause a crash.
• Improved the manual.
• Fixed an issue in the presolve ocurring very rarely. 
• Fixed a bug causing an assert on very large problems.
• FLEXlm has been upgraded to version 11.9.1 because some bugs in FLEXlm has been fixed. 
• Fixed several issues appearing when the input data that had NANs.
• Fixed a bug in the conic optimizer causing an assert for certain problems. 
• Fixed a bug in MSK_analyzeproblem that could cause a segmentation fault on problems containing cones.

Portfolio optimization whitepaper

We revised our whitepaper on portfolio optimization using MATLAB.

The whitepaper gives an introduction to portfolio optimization using the MOSEK conic optimizer from MATLAB, and includes topics such as:

• Conic formulations of standard Markowitz portfolio problems.
• Minimum risk/maximum return formulations.
• Computing the efficient frontier.
• Computing the maximum Sharpe ratio.
• Exploiting low-rank structure in the covariance matrix to reduce solution time, including factor models.
• Transaction costs with market impact, modeled as a conic quadratic problem.
• Transactions cost with a fixed term, modeled as a conic mixed-integer problem. 

Furthermore, the revised version uses a dataset with historical data for the S&P500 index in all the examples.

The whitepaper can be downloaded from the MOSEK publications page.

MOSEK is sponsoring 3 optimization conferences.

MOSEK is happy to sponsor the 2011 conferences:

• MOPTA 2011.
• CPAIOR 2011. 8th International Conference on Integration of Artificial Intelligence and Operations Research. 
• SIAM Conference on Optimization (OP11).

MOSEK will also have two presentations at the SIAM optimization conference.