stdgrb: Python Gurobi solvers for standard problems

This is a fast cython wrapper for strandard dense optimization that calls gurobi through its C interface. It is more efficient for dense problems than the gurobi python interface that cannot accept directly numpy arrays for constraints.

Install

The only strong dependencies are

• numpy
• cython (for the compilation)

In order to build install stdgrb you need to define first environement variables GUROBI_HOME and GUROBI_VERSION so that the compiler can find the include files and link the good library.

this can be done with

export GUROBI_HOME=/path/to/gurobi/linux64
export GUROBI_VERSION=gurobi75 # for gurobi 7.5

you can find the gurobi version in the GUROBI_HOME/lib folder by seeking the libgurobi[XX].so and setting GUROBI_VERSION=gurobi[XX] (be carefull to remove the 'lib').

When the environement variables are set, you can install the module with

python setup.py install # --user if local install

In you want to compile stdgrb inplace execute the following command

python setup.py build_ext --inplace

This will allow execution of example scripts directly in the stdgrb folder.

Use the module

Linear program (LP,ILP,MILP) lp_solve

lp_solve(c,A=None,b=None,lb=None,ub=None,nbeq=0,typevar=None, method=-1,logtoconsole=1, crossover=-1)

Solves the following optimization problem:

with optional equality constraints and integer variables in x.

Parameters:

• c : (d,) ndarray, float64,
  Linear cost vector
• A : (n,d) ndarray, float64, optional,
  Linear constraint matrix
• b : (n,) ndarray, float64, optional,
  Linear constraint vector
• lb : (d) ndarray, float64, optional,
  Lower bound constraint
• ub : (d) ndarray, float64, optional,
  Upper bound constraint
• nbeq: int, optional,
  Treat the nbeq first lines of A as equality constraints.
• typevar: int, (d,) ndarray, int, optional
  Integer or array of integer defining the type of variables (default
  continuous).
  Select typevar from:
  • 0 : Continuous
  • 1 : Binary
  • 2 : Integer
  • 3 : Semi-continuous (0.0 or value in const)
  • 4 : Semi-integer (0 or value in const)
• method : int, optional,
  Selected solver from (see gurobi documentation)
  • -1=automatic (default),
  • 0=primal simplex,
  • 1=dual simplex,
  • 2=barrier,
  • 3=concurrent,
  • 4=deterministic concurrent,
  • 5=deterministic concurrent simplex
• logtoconsole : int, optional,
  If 1 the print log in console,
• crossover : int, optional,
  Select crossover strategy for interior point (see gurobi documentation)

Returns:

• x: (d,) ndarray,
  Optimal solution x
• val: float,
  Optimal value of the objective (None if optimization error)

Quadratic program (QP,IQP,MIQP) qp_solve

qp_solve(Q,c=None,A=None,b=None,lb=None,ub=None,nbeq=0,typevar=None, method=-1,logtoconsole=1, crossover=-1)

Solves the following optimization problem:

with optional equality constraints and integer variables in x.

Parameters:

• Q : (d,d) ndarray, float64, optional
  Quadratic cost matrix matrix
• c : (d,) ndarray, float64, optional,
  Linear cost vector
• A : (n,d) ndarray, float64, optional,
  Linear constraint matrix
• b : (n,) ndarray, float64, optional,
  Linear constraint vector
• lb : (d) ndarray, float64, optional,
  Lower bound constraint
• ub : (d) ndarray, float64, optional,
  Upper bound constraint
• nbeq: int, optional,
  Treat the nbeq first lines of A as equality constraints.
• typevar: int, (d,) ndarray, int, optional
  Integer or array of integer defining the type of variables (default
  continuous).
  Select typevar from:
  • 0 : Continuous
  • 1 : Binary
  • 2 : Integer
  • 3 : Semi-continuous (0.0 or value in const)
  • 4 : Semi-integer (0 or value in const)
• method : int, optional,
  Selected solver from (see gurobi documentation)
  • -1=automatic (default),
  • 0=primal simplex,
  • 1=dual simplex,
  • 2=barrier,
  • 3=concurrent,
  • 4=deterministic concurrent,
  • 5=deterministic concurrent simplex
• logtoconsole : int, optional,
  If 1 the print log in console,
• crossover : int, optional,
  Select crossover strategy for interior point (see gurobi documentation)

Returns:

• x: (d,) ndarray,
  Optimal solution x
• val: float,
  Optimal value of the objective (None if optimization error)