Try our new documentation site (beta).


dense_cs.cs


/* Copyright 2020, Gurobi Optimization, LLC */

/* This example formulates and solves the following simple QP model:

     minimize    x + y + x^2 + x*y + y^2 + y*z + z^2
     subject to  x + 2 y + 3 z >= 4
                 x +   y       >= 1
                 x, y, z non-negative

   The example illustrates the use of dense matrices to store A and Q
   (and dense vectors for the other relevant data).  We don't recommend
   that you use dense matrices, but this example may be helpful if you
   already have your data in this format.
*/

using System;
using Gurobi;

class dense_cs {

  protected static bool
    dense_optimize(GRBEnv    env,
                   int       rows,
                   int       cols,
                   double[]  c,      // linear portion of objective function
                   double[,] Q,      // quadratic portion of objective function
                   double[,] A,      // constraint matrix
                   char[]    sense,  // constraint senses
                   double[]  rhs,    // RHS vector
                   double[]  lb,     // variable lower bounds
                   double[]  ub,     // variable upper bounds
                   char[]    vtype,  // variable types (continuous, binary, etc.)
                   double[]  solution) {

    bool success = false;

    try {
      GRBModel model = new GRBModel(env);

      // Add variables to the model

      GRBVar[] vars = model.AddVars(lb, ub, null, vtype, null);

      // Populate A matrix

      for (int i = 0; i < rows; i++) {
        GRBLinExpr expr = new GRBLinExpr();
        for (int j = 0; j < cols; j++)
          if (A[i,j] != 0)
            expr.AddTerm(A[i,j], vars[j]); // Note: '+=' would be much slower
        model.AddConstr(expr, sense[i], rhs[i], "");
      }

      // Populate objective

      GRBQuadExpr obj = new GRBQuadExpr();
      if (Q != null) {
        for (int i = 0; i < cols; i++)
          for (int j = 0; j < cols; j++)
            if (Q[i,j] != 0)
              obj.AddTerm(Q[i,j], vars[i], vars[j]); // Note: '+=' would be much slower
        for (int j = 0; j < cols; j++)
          if (c[j] != 0)
            obj.AddTerm(c[j], vars[j]); // Note: '+=' would be much slower
        model.SetObjective(obj);
      }

      // Solve model

      model.Optimize();

      // Extract solution

      if (model.Status == GRB.Status.OPTIMAL) {
        success = true;

        for (int j = 0; j < cols; j++)
          solution[j] = vars[j].X;
      }

      model.Dispose();

    } catch (GRBException e) {
      Console.WriteLine("Error code: " + e.ErrorCode + ". " + e.Message);
    }

    return success;
  }

  public static void Main(String[] args) {
    try {
      GRBEnv env = new GRBEnv();

      double[] c = new double[] {1, 1, 0};
      double[,] Q = new double[,] {{1, 1, 0}, {0, 1, 1}, {0, 0, 1}};
      double[,] A = new double[,] {{1, 2, 3}, {1, 1, 0}};
      char[] sense = new char[] {'>', '>'};
      double[] rhs = new double[] {4, 1};
      double[] lb = new double[] {0, 0, 0};
      bool success;
      double[] sol = new double[3];

      success = dense_optimize(env, 2, 3, c, Q, A, sense, rhs,
                               lb, null, null, sol);

      if (success) {
        Console.WriteLine("x: " + sol[0] + ", y: " + sol[1] + ", z: " + sol[2]);
      }

      // Dispose of environment
      env.Dispose();

    } catch (GRBException e) {
      Console.WriteLine("Error code: " + e.ErrorCode + ". " + e.Message);
    }
  }
}

Try Gurobi for Free

Choose the evaluation license that fits you best, and start working with our Expert Team for technical guidance and support.

Evaluation License
Get a free, full-featured license of the Gurobi Optimizer to experience the performance, support, benchmarking and tuning services we provide as part of our product offering.
Academic License
Gurobi supports the teaching and use of optimization within academic institutions. We offer free, full-featured copies of Gurobi for use in class, and for research.
Cloud Trial

Request free trial hours, so you can see how quickly and easily a model can be solved on the cloud.

Search

Gurobi Optimization