2.1.6 Termination Criteria

The iterations are continued until convergence which is denoted by either a small relative change in the sum of squares error (F), or by a small change in each parameter value relative to the previous iteration values. Other termination criteria are indicative of error conditions that occur when the optimization problem is ill posed such as when:

• The number of iterations exceeds a maximum number: .
• The conditioning factor exceeds a preset maximum value: . In this case, the algorithm is failing to move forward even though the method of steepest descent is used.
• The norm of the gradient is very small. This indicates the failure to find a search direction: .

As in other optimization problems, the convergence to a solution does not guarantee that the global minimum of F has been reached.