How can you use Gurobi's vbasis and cbasis with the R-Interface?

问题

I found an explanation in the docs here and here:

vbasis
Variable basis status values for the computed optimal basis. You generally should not concern yourself with the contents of this array. If you wish to use an advanced start later, you would simply copy the vbasis and cbasis arrays into the corresponding fields for the next model. This array contains one entry for each column of A.

cbasis
Constraint basis status values for the computed optimal basis. This array contains one entry for each row of A.

And later on:

Finally, if the final solution is a basic solution (computed by simplex), then vbasis and cbasis will be present.

I don't understand why I don't get those values - maybe I miss something.

Reproducible example:

model <- list()
model$A          <- matrix(c(1, 1, 1, 0, 
                             1, 0, 0, 0, 
                             0, 1, 0, 0, 
                             0, 0, 1, 0, 
                             0, 0, 0, 1), nrow = 5, ncol = 4, byrow = T)
model$obj        <- c(2, -5, 3, 10)
model$modelsense <- "min"
model$rhs        <- c(15, 7, 3, 5, 1)
model$sense      <- c('=', '<=', '<=', '<=', '>')
model$vtype      <- 'I'
params <- list(OutputFlag = 1, Presolve = 2, TimeLimit = 3600)
result <- gurobi::gurobi(model, params)

names(result)
[1] "status" "runtime" "itercount" "baritercount" "nodecount" "objval" "x" "slack" "objbound"

回答1:

Thanks to Greg Glockner's answer here I found something in the documentation which I copy here.

It seems quite a lot has improved with Gurobi version 8.0.0!

# Copyright 2018, Gurobi Optimization, LLC
#
# This example reads an LP model from a file and solves it.
# If the model can be solved, then it finds the smallest positive variable,
# sets its upper bound to zero, and resultolves the model two ways:
# first with an advanced start, then without an advanced start
# (i.e. 'from scratch').

library(Matrix)
library(gurobi)

args <- commandArgs(trailingOnly = TRUE)
if (length(args) < 1) {
  stop('Usage: Rscript lpmod.R filename\n')
}

# Read model
cat('Reading model',args[1],'...')
model <- gurobi_read(args[1])
cat('... done\n')

# Determine whether it is an LP
if ('multiobj'  %in% names(model) ||
    'sos'       %in% names(model) ||
    'pwlobj'    %in% names(model) ||
    'cones'     %in% names(model) ||
    'quadcon'   %in% names(model) ||
    'genconstr' %in% names(model)   ) {
  stop('The model is not a linear program\n')
}

# Detect set of non-continuous variables
intvars    <- which(model$vtype != 'C')
numintvars <- length(intvars)
if (numintvars > 0) {
  stop('problem is a MIP, nothing to do\n')
}

# Optimize
result <- gurobi(model)
if (result$status != 'OPTIMAL') {
  cat('This model cannot be solved because its optimization status is',
      result$status, '\n')
  stop('Stop now\n')
}

# Recover number of variables in model
numvars   <- ncol(model$A)

# Ensure bounds array is initialized
if (is.null(model$lb)) {
  model$lb <- rep(0, numvars)
}
if (is.null(model$ub)) {
  model$ub <- rep(Inf, numvars)
}

# Find smallest (non-zero) variable value with zero lower bound
x      <- replace(result$x, result$x < 1e-4, Inf)
x      <- replace(x, model$lb > 1e-6, Inf)
minVar <- which.min(x)
minVal <- x[minVar]

# Get variable name
varname <- ''
if (is.null(model$varnames)) {
  varname <- sprintf('C%d',minVar)
} else {
  varname <- model$varnames[minVar]
}

cat('\n*** Setting', varname, 'from', minVal, 'to zero ***\n\n')
model$ub[minVar] <- 0

# Set advance start basis information
model$vbasis <- result$vbasis
model$cbasis <- result$cbasis

result2   <- gurobi(model)
warmCount <- result2$itercount
warmTime  <- result2$runtime

# Reset-advance start information
model$vbasis <- NULL
model$cbasis <- NULL

result2   <- gurobi(model)
coldCount <- result2$itercount
coldTime  <- result2$runtime

cat('\n*** Warm start:', warmCount, 'iterations,', warmTime, 'seconds\n')
cat('\n*** Cold start:', coldCount, 'iterations,', coldTime, 'seconds\n')

# Clear space
rm(model, result, result2)

来源：https://stackoverflow.com/questions/49746686/how-can-you-use-gurobis-vbasis-and-cbasis-with-the-r-interface

标签

gurobi