13.07.2015 Views

A Proposal of a Mathematical Roadmap for Scilab DRAFT - Projects

A Proposal of a Mathematical Roadmap for Scilab DRAFT - Projects

A Proposal of a Mathematical Roadmap for Scilab DRAFT - Projects

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

• the Taucs package, which provides a direct Cholesky algorithm <strong>for</strong> sparse matrices: http://www.scilab.org/product/man/taucs_chfact.html• the UMFPACK package, which provides a direct LU algorithm <strong>for</strong> sparse matrices: http://www.scilab.org/product/man/umf_lufact.htmlThe Benchmark shown at : scilab/modules/umfpack/examples/bench.txt showsthat the UMFPACK functions are much faster than the backslash operator from theSparse package.The scilin project : http://www.irisa.fr/aladin/codes/SCILIN/ aims at providingsparse linear algebra routines. This project is currently under update on<strong>Scilab</strong>’s Forge : http://<strong>for</strong>ge.scilab.org/index.php/p/scilin/Some routines developped in Scilin have already been integrated into <strong>Scilab</strong> :• pcg : precondioned conjugate gradient• qml : quasi minimal resiqual method with preconditioningThere are some questions which are left to be determined :• what is the purpose <strong>of</strong> spchol (sparse cholesky factorization) : what is thelibrary used by spchol ?The current state <strong>of</strong> <strong>Scilab</strong> with respect to sparse linear algebra is confusing inthe sense <strong>of</strong> many tools are providing nearly similar features (like lufact / umf lufact,spchol/taucs chfact), or are iterative solvers (pcg, qmr) using preconditionners whichare not provided by <strong>Scilab</strong>.4 OptimizationThis is a list <strong>of</strong> missing optimization features in <strong>Scilab</strong>.available in current <strong>Scilab</strong> toolboxes.Theses features are not• Non-linear objective with non linear constraints4.1 Why it fits in <strong>Scilab</strong> scopeOptimization features are already part <strong>of</strong> current <strong>Scilab</strong> features and several <strong>Scilab</strong>toolboxes.4.2 Existing features in MatlabMatlab provides the fmincon function, which searches <strong>for</strong> the local minimum <strong>of</strong> anonlinear function with nonlinear constraints : fminconMatlab provides the bintprog, which solves binary integer programming problems: bintprog13

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

Saved successfully!

Ooh no, something went wrong!