-
Notifications
You must be signed in to change notification settings - Fork 1
Exclude Lpsolve
Lpsolve is a C package (with R interface) to solve Linear Programs (LPs). GeomaScale's C++ library volesti
uses the C code of Lpsolve to solve LPs. These computations are useful to compute the maximum interior ball
of a convex polytope. However, for several high-dimensional polytopes, Lpsolve fails to compute the inner
ball correctly. This leads to an unexpected halt of volesti
.
The project will exclude from the R, C++ interfaces of volesti
the Lpsolve solver and replace it
with other R and C++ software that can solve LPs.
A few CRAN R packages that can solve LPs are: (i) nloptr, (ii) quadprog, (iii) qpmadr, (iv) ipop.
However, the student will have to take into account other open-source R and C++ packages.
See also https://github.com/GeomScale/dingo/issues/84
The student will have (i) to make research to find the most CRAN optimization packages that can solve LPs, (ii) repeat step (i) for C++ and python software, and (iii) to perform
extended comparison experiments for the software in (i) and (ii), and (iii) implement the computation of the maximum inner ball of a polytope
in R and C++. Then, the contributor will have to integrate the new function to the python, R and C++ interface of volesti
.
Difficulty: Easy
Medium (175 hours)
- Required: C++, R, python
- Preferred: Experience with mathematical software is a plus
This is a very useful project for volesti
, since it could be used for more complex inputs without failing to perform
the preprocess that is necessary for sampling and volume computation.
-
Apostolos Chalkis <tolis.chal at gmail.com> is a Research Engineer at Quantagonia GmbH. He is an expert in statistical software, computational geometry, and optimization, and has previous GSoC student experience (2018 & 2019) and mentoring experience with GeomScale (from 2020 to 2023).
-
Manolis Christoforou is an expert in open-source development, machine learning, and optimization.
-
Zafeirakis Zafeirakopoulos is an expert in implementing and benchmarking geometric and algebraic algorithms and has previous GSOC experience with the R-project (2018, 2019).
Students, please contact the first and the third mentor after completing at least one of the tests below.
Students, please do one or more of the following tests before contacting the mentors above.
- Easy: Built and run the R and C++ interfaces of
volesti
. Compute the maximum ball of a polytope using the existing functions. - Medium: Report one more CRAN package except for nloptr, quadprog, solveqp, and ipop that can compute the maximum ball of a polytope. Report also 2 open-source C++ packages that solve linear programs.
- Hard: Implement the maximum ball computation with CRAN package
nloptr
.
For tips and references contact the Mentors!
Students, please post a link to your test results here.
- EXAMPLE STUDENT 1 NAME, LINK TO GITHUB PROFILE, LINK TO TEST RESULTS.