Skip to contents

Propagates the row and column bounds of an LP to a fixed point and returns the LP with the implied interval of each column, lo_impl and up_impl, on its column index. The interval contains every feasible value of the variable; it is tight only where a single row decides it.

Usage

lp_bounds(lp, objective = NULL, max_pass = 50L, tol = 1e-09, verbose = FALSE)

Arguments

lp

An LP from model_to_lp() or lp_presolve().

objective

A known feasible objective value, or NULL. When given, the objective row sum c_j x_j <= objective (>= for a maximisation) takes part, which bounds variables that rows alone leave open.

max_pass

Passes over the rows; propagation stops earlier when no bound moves by more than tol.

tol

Minimal improvement of a bound that counts as a change.

verbose

Report each pass.

Value

The LP with lo_impl, up_impl on col_index and a bounds element: passes, tightened (columns whose interval shrank), fixed (zero-width intervals), open (an infinite end left) and infeasible (a row whose activity range misses its bounds, by index).

Examples

lp <- model_to_lp(example_models$toy)
b <- lp_bounds(lp)
b$bounds[c("tightened", "fixed", "open")]
#> $tightened
#> [1] 1
#> 
#> $fixed
#> [1] 0
#> 
#> $open
#> [1] 13
#> 
head(b$col_index[, c("symbol", "lo", "up", "lo_impl", "up_impl")])
#>    symbol    lo    up lo_impl up_impl
#>    <char> <num> <num>   <num>   <num>
#> 1:   vCap     0   Inf       0     Inf
#> 2:   vCap     0   Inf       0     Inf
#> 3:   vCap     0   Inf       0     Inf
#> 4:   vCap     0   Inf       0     Inf
#> 5:   vCap     0   Inf       0     Inf
#> 6:   vCap     0   Inf       0     Inf