clingo[LP] extends the ASP solver clingo with linear constraints as dealt with in Linear Programming (LP).
Install via conda:
conda install -c potassco -c conda-forge clingo-lpThe clingoLP propagator requires an LP solver. The default is lpsolve but cplex can be used optionally.
IBM also provides a promotional version of cplex that is sufficient to solve small problems limited to 1000 variables and 1000 constraints.
conda install -c ibmdecisionoptimization cplex`To solve larger problems, you need to use the full version of CPLEX Studio.
LP constraints can be expressed as follows:
| LP constraints | ClingoLP Syntax |
|---|---|
| w1x1+...+wnxn >= k | &sum{w1*x1;...;wn*xn} >= k |
| domain(x)={l,...,u} | &dom{l..u} = x |
| maximize: w1x1+...+wnxn | &maximize{w1*x1;...;wn*x1} (minimize analogous) |
To avoid syntax clashes, you must quote " real numbers. Instead of 1.5 write "1.5".
clingoLP [number] [options] [files]
Options:
--lp-solver=<arg> : Set LP solver
<arg>: {lps,cplx} (default lp-solver=lps)
--[no-]show-lp-solution : Show LP solution and value of objective function
--accuracy=n : Set decimal position of LP solver accuracy (default accuracy=3)
--epsilon=n,m : Set epsilon to convert lhs > k into lhs >= k+n*10^-m (default epsilon=1,3)
--[no-]trace : Enables detailed output of theory propagation
--core-confl=n : Searches for core conflicts if at least n% of the theory atoms are decided
(default core-confl=20)
--prop-heur=n : Starts a solve call of the LP solver if at least n% of the theory atoms are decided
(default prop-heur=0)
--[no-]ilp : Sets the LP solver to solve an Integer Linear Programming (ILP) problemExample:
clingoLP 0 --show-lp-solution example_encoding.lp example_instance.lpFor more options you can ask for help as follows:
clingoLP --help