21 Jan

SolverStudio, OpenSolver and Bioinformatics

Came across the following quote in Briefings in Bioinformatics. “Interestingly, Microsoft Excel (http://office.microsoft.com/en-us/excel/), the commonly used spreadsheet package for storing metabolic models, has all the necessary components to create and analyse FBA models including well-designed form-based interface, plotting and drawing facilities, in-built optimization solvers and scripting language for automating tasks (i.e. MACROS). It should be noticed that Excel-based FBA application may face technical challenges in handling large-size models due to the limitation in its in-built optimization solver. However, this issue can be appropriately resolved by using relevant software technologies such as OpenSolver (http://opensolver.org), an open-source optimization solver for Excel that runs on advanced COIN-OR CBC optimization engine (https://projects.coin-or.org/Cbc) and SolverStudio (http://solverstudio.org), a software framework that can integrate Excel with other open sources as well as commercial solvers, e.g. GLPK (http://www.gnu.org/software/glpk/), COIN CLP (http://www.coin-or.org/Clp/), CPLEX and GUROBI (http://www.gurobi.com/).” Is this an area we OR types need to be looking at more seriously?

On another matter, thanks, Bob, for mentioning SolverStudio in your AMPL talk at Informs.

28 Sep

NEOS: Cloud Optimisation using Excel

We’ve been working on improving SolverStudio to meet the needs of a student class we are teaching. As part of these changes, we are currently beta testing a new version of SolverStudio that allows AMPL models to be solved “in the cloud” using the excellent NEOS on-line optimisation service.

SolverStudio can now take an AMPL model, combine it with the data on your spreadsheet, and send it to NEOS to solve. The results are then sent back to your computer, and appear in your spreadsheet. Using NEOS instead of a local copy of AMPL is as simple as changing the model language from AMPL to ‘AMPL on NEOS’. NEOS supports a wide range of solvers, making it very easy to experiment with different ones, including heuristics.

This is all possible thanks to the XMLRPC (XML remote procedue call) support that NEOS provides. NEOS even provide a Python example, which made it easy to add a new ‘AMPLNEOS’ language to SolverStudio and then write the Python support files to pass the model and data to and from NEOS. Thanks, NEOS, for an excellent service.