CVXR
Disciplined Convex Optimization
An object-oriented modeling language for disciplined
convex programming (DCP) as described in Fu, Narasimhan, and Boyd
(2020,
CVXR provides an object-oriented modeling language for convex
optimization, similar to CVX, CVXPY, YALMIP, and Convex.jl. It
allows the user to formulate convex optimization problems in a natural
mathematical syntax rather than the restrictive standard form required
by most solvers. The user specifies an objective and set of constraints
by combining constants, variables, and parameters using a library of
functions with known mathematical properties. CVXR then applies signed
disciplined convex programming
(DCP) to
verify the problem’s convexity. Once verified, the problem is converted
into standard conic form using graph implementations and passed to a
cone solver such as ECOS or
SCS.
CVXR includes several open source solvers in addition to the default ECOS and SCS. Recent versions (0.97+) also include support for commercial solvers such as MOSEK and GUROBI.
Installation
This package is now released on CRAN, so you can install the current
released version as you would any other package for R, version 3.4 and
higher. (CVXR is known to work with earlier versions of R too, but we
don’t check our releases against older versions of R.)
install.packages('CVXR', repos = "https://CRAN.R-project.org")
Development versions can be installed from the Github repository assuming you have the development tools for R available, including the C compilers etc. Execute:
library(devtools)install_github("anqif/CVXR")
Tutorial
A number of tutorial examples are available on the CVXR website.
News
CVXR 0.99-5
- Bug fix for LogSumExp atom. This should address the issue reported on StackOverflow
- Require ECOSolveR version 0.5.1 and above to avoid convolution example failure on 32-bit platforms.
CVXR 0.99-4
-
Updated tests to check against
scsversion 1.2.3, which now returnsoptimalrather thanoptimal_inaccurrateas it did earlier (test-g01-non_optimal.R#35). -
Some cleanup of manual tests. New SCS returns slightly different values for
Sigmain manual test (test-vignette#646) for worst-case covariance example.
CVXR 0.99-3
- Bug fix to MOSEK and GUROBI solver interfaces; dimensions now converted to expected python types.
- Sanitized src directory: proper naming of CVXR header to include while compiling attributes with Rcpp.
CVXR 0.99-2
- Typo fixes to URLs.
CVXR 0.99-1
- Bug fix: Updated python glue for both version 2 and 3
(
gurobiglue.py) - Bug fix: Workaround for zero extent sparse matrices in R not being
handled by reticulate (
mosekglue.py)
CVXR 0.99
- Bug fix: duplicated integer and boolean indices.
- Bug fix: correct typo in constraint specification to GLPK.
- Added tutorial articles based on v0.99 to CVXR website on using other solvers, integer programming, MOSEK and GUROBI examples.
CVXR 0.98-1
- Minor typographical fixes.
CVXR 0.98
-
Dropped
delay_loadparameter dropped inreticulate::import_from_path, per changes inreticulate. -
Cleaned up hooks into reticulate for commercial solvers.
CVXR 0.97-1
- Minor typo and documentation fixes.
CVXR 0.97
- Added
LPSOLVEvialpSolveAPI - Added
GLPKviaRglpk - Added
MOSEK - Added
GUROBI - Bug fix: issue #25.
All CVXR expressions retain dimensions. Culprit was
drop = FALSE(in functionIndex.get_special_slice) as suspected.
CVXR 0.96
-
Added a note that CVXR can probably be compiled from source for earlier versions of R. This is issue #24
-
Bug fix: issue #28 Function
intf_sign(interface.R) was unnecessarily using a tolerance parameter, now eliminated.
CVXR 0.95
-
Updated Solver.solve to adapt to new ECOSolveR. Require version 0.4 of ECOSolveR now.
-
Updated
unpack_resultsto behave exactly like in CVXPY. Added documentation and testthat tests. Documented in Getting Faster Results article.
CVXR 0.94-4
- First CRAN release 2017-11-20.
CVXR pre-0.94-4
- Several wrong turns and much hand-wringing. Complete rewrite in preparation for release.