Each factor is linked to a series of values between 0 and 1 that stand for the tolerability of non-ideal values. These values are called tolerance options.Tolerance options become a tolerance hypothesis when used. Tolerance hypothesis were created in order to verify the validity of this hypothesis. A tolerance test vector is the combination of values that represent the tolerance in each factor.
The FACTORFINDER-Procedure recombines new tolerance test vectors from single tolerance values. Information on adequate combinations of tolerance values are indirectly recorded in the counter matrix. The counter matrix saves information on the tolerance options that are most likely to be successful.
The probability of each tolerance option to be part of future tolerance test vectors depends on its proportion in the sum of all counter values of the respective factor. By this, new tolerance hypothesis with random variations for each factor are created. Through random influence rare tolerance options that are represented by small counter values find their way into the tolerance test vector.
A counter matrix may be the source of any number of competing tolerance test vectors. From all tolerance options of a factor only one option can be selected. The chance of being chosen reflects former successful usages and its fitness to express the tolerability of non-ideal values in the current candidate example. At the same time, not only the primary task is reflected in this selection but also random influences to generate new solutions.
At the moment of test, the test vector is applied to the example base und afterwards all its elements again separated. Depending on the result of the test the counter values are reduced or enlarged. A test vector reflects information that is not only from its direct predecessor. Instead, the procedure disassembles each test vector completely. In each factor, it is marked whether the current tolerance hypothesis has been successfully tested. In this case, the counter that belongs to that tolerance option is enlarged.
The procedure does not save complete sequences of successful tolerance test vectors. The counter matrix expresses information on the interaction of the factors. Successful combinations of tolerance options are recorded by means of higher counter values of the respective tolerance options. Implicitly, each counter value contains information on the other factors and on the results of previous procedure cycles. The distribution of the counter values results from their adequacy in respect to all the examples, in respect to the target criterion and the tolerance hypothesis for the remaining factors.