ECOS_csolve {ECOSolveR} | R Documentation |
Solve a conic optimization problem
Description
The function ECOS_csolve
is a wrapper around the ecos
csolve
C function. Conic constraints are specified using the
and
parameters and can be
NULL
and zero
length vector respectively indicating an absence of conic
constraints. Similarly, equality constraints are specified via
and
parameters with
NULL
and empty vector
values representing a lack of such constraints. At most one of the
pair or
is allowed to be absent.
Usage
ECOS_csolve(
c = numeric(0),
G = NULL,
h = numeric(0),
dims = list(l = integer(0), q = NULL, e = integer(0)),
A = NULL,
b = numeric(0),
bool_vars = integer(0),
int_vars = integer(0),
control = ecos.control()
)
Arguments
c |
the coefficients of the objective function; the length of
this determines the number of variables |
G |
the inequality constraint matrix in one of three forms: a
plain matrix, simple triplet matrix, or compressed column
format, e.g. dgCMatrix-class. Can also be
|
h |
the right hand size of the inequality constraint. Can be empty numeric vector. |
dims |
is a list of three named elements: |
A |
the optional equality constraint matrix in one of three
forms: a plain matrix, simple triplet matrix, or compressed
column format, e.g. dgCMatrix-class. Can be
|
b |
the right hand side of the equality constraint, must be
specified if |
bool_vars |
the indices of the variables, 1 through |
int_vars |
the indices of the variables, 1 through |
control |
is a named list that controls various optimization parameters; see ecos.control. |
Value
a list of 8 named items
- x
primal variables
- y
dual variables for equality constraints
- s
slacks for
,
- z
dual variables for inequality constraints
- infostring
gives information about the status of solution
- retcodes
a named integer vector containing four elements
- exitflag
0=
ECOS_OPTIMAL
, 1=ECOS_PINF
, 2=ECOS_DINF
, 10=ECOS_INACC_OFFSET
, -1=ECOS_MAXIT
, -2=ECOS_NUMERICS
, -3=ECOS_OUTCONE
, -4=ECOS_SIGINT
, -7=ECOS_FATAL
. See ECOS_exitcodes
.
- iter
the number of iterations used
- mi_iter
the number of iterations for mixed integer problems
- numerr
a non-zero number if a numeric error occurred
- summary
a named numeric vector containing
- pcost
value of primal objective
- dcost
value of dual objective
- pres
primal residual on inequalities and equalities
- dres
dual residual
- pinf
primal infeasibility measure
- dinf
dual infeasibility measure
- pinfres
primal infeasibility residual
- dinfres
dual infeasibility residual
- gap
duality gap
- relgap
relative duality gap
- r0
Unknown at the moment to this R package maintainer.
- timing
a named numeric vector of timing information consisting of
- runtime
the total runtime in ecos
- tsetup
the time for setup of the problem
- tsolve
the time to solve the problem
Details
A call to this function will solve the problem:
minimize , subject to
, and
.
Variables can be constrained to be boolean (1 or 0) or integers. This is indicated
by specifying parameters bool_vars
and/or int_vars
respectively. If so
indicated, the solutions will be found using a branch and bound algorithm.
Examples
## githubIssue98
cat("Basic matrix interface\n")
Gmat <- matrix(c(0.416757847405471, 2.13619609566845, 1.79343558519486, 0, 0,
0, 0, -1, 0, 0, 0, 0.056266827226329, -1.64027080840499, 0.841747365656204,
0, 0, 0, 0, 0, -1, 0, 0, 0, 0, 0, 0.416757847405471, 2.13619609566845,
1.79343558519486, 0, 0, 0, -1, 0, 0, 0, 0, 0.056266827226329, -1.64027080840499,
0.841747365656204, 0, 0, 0, 0, -1, 0, 0, 0, 0, 0, 0, -1, 0, 0, 0, 0), ncol = 5L)
c <- as.numeric(c(0, 0, 0, 0, 1))
h <- as.numeric(c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0))
dims <- list(l = 6L, q = 5L, e = 0L)
ECOS_csolve(c = c, G = Gmat, h = h,
dims = dims,
A = NULL, b = numeric(0))
cat("Simple Triplet Matrix interface, if you have package slam\n")
if (requireNamespace("slam")) {
ECOS_csolve(c = c, G = slam::as.simple_triplet_matrix(Gmat), h = h,
dims = dims,
A = NULL, b = numeric(0))
}
if (requireNamespace("Matrix")) {
ECOS_csolve(c = c, G = Matrix::Matrix(Gmat), h = h,
dims = dims,
A = NULL, b = numeric(0))
}
## Larger problems using saved data can be found in the test suite.
## Here is one
if (requireNamespace("Matrix")) {
MPC01 <- readRDS(system.file("testdata", "MPC01_1.RDS", package = "ECOSolveR"))
G <- Matrix::sparseMatrix(x = MPC01$Gpr, i = MPC01$Gir, p = MPC01$Gjc,
dims = c(MPC01$m, MPC01$n), index1 = FALSE)
h <- MPC01$h
dims <- lapply(list(l = MPC01$l, q=MPC01$q, e=MPC01$e), as.integer)
retval <- ECOS_csolve(c = MPC01$c, G=G, h = h, dims = dims, A = NULL, b = NULL,
control = ecos.control(verbose=1L))
retval$retcodes
retval$infostring
retval$summary
}