Embed returns the outcome of the optimization process through the value of inform in the summary of the report. All possible values of inform are described below. When the results are not what you expected, you will need to troubleshoot the diagram.

Inform |
Meaning | |

0 |
Kuhn-Tucker conditions satisfied. This is the best possible indicator that an optimal point has been found. | |

1 |
Fractional change in objective less than epstop for nstop consecutive iterations. This is not as good as inform = 0, but still indicates the likelihood that an optimal point has been found. | |

2 |
All remedies have failed to find a better point. You should check cost and constraint functions and bounds for consistency and, perhaps, try other starting values. | |

3 |
Number of completed 1D searches exceeded limser. You should check cost and constraint functions and bounds for consistency and, perhaps, try other starting values. | |

4 |
Objective function is unbounded. Embed has observed dramatic change in the objective function over several steps. This is a good indication that the objective function is unbounded. If this is not the case, you should check the cost and constraint functions and bounds for consistency. | |

5 |
Feasible point not found. Embed was not able to find a feasible point. If the problem is believed to be feasible, you should check the cost and constraint functions and bounds for consistency and, perhaps, try other starting values. | |

6 |
Degeneracy has been encountered. The point returned may be close to optimal. You should check the cost and constraint functions and bounds for consistency and, perhaps, try other starting values. | |

7 |
Noisy and non-smooth function values. Possible singularity or errors in the cost and constraint function evaluations. The solution returned may be as accurate as possible with the current simulation set-up. It may be necessary to increase the accuracy of the cost and constraint functions by reducing the step size set in the System Properties dialog box. | |

8 |
The optimization process was terminated by a user request through the Stop command. | |

9 |
Maximum number of simulation runs exceeded. The number of simulation runs is set under Max Iterations in the Optimization Properties dialog box. Set Max Iterations to 0 if you want Embed to continue to run until it converges to a solution or until you explicitly stop it by choosing the Stop command. | |

-1 |
Fatal Error. Some condition, such as number of variables ≤ 0, was encountered. Embed documented the condition in the report and terminated. In this case, you need to correct the Embed diagram and rerun Embed. | |

Messages inform = 0, inform = 1, and inform = 2 are by far the most common. Message inform = 0 implies the highest level of confidence that at least a local optimum has been found; message inform = 1 implies less confidence; and message inform = 2 even less.

In message inform = 0, the Kuhn-Tucker conditions are first-order necessary conditions that hold if the current point is at least a local optimum and all functions have continuous first partial derivatives.

In message inform = 2, the following sequence of events has occurred:

• No improved point was located along the last search direction.

• Change of basis was attempted (if one had not already been done).

• If the search direction was not the negative reduced gradient, this direction is tried.

• If any variables with values at a bound have reduced gradient components indicating that releasing them from that bound could improve the objective, one such variable is allowed to leave its bound.

In other words, Embed tried all known remedies, and none of these remedies improved the objective function, so the program terminated.

Regardless of which of termination messages inform = 0, 1, and 2 is returned, the current point may be (nearly) optimal. Message inform = 0 may fail to appear because the variables or constraints of the problem are poorly scaled.

Message inform = 5 is returned when Phase I terminates and the final point is not feasible. In this case, there may be no point satisfying all problem constraints. If this is not the believed to be the case, the user can select new starting values for the parameterUnknowns and rerun.

If you are unsatisfied with the solution found by Embed, you may need to troubleshoot the diagram.