Rule Window

The Rule Window is or fuzzy rule base specification. Rules are built using input and output labels which must have been previously defined. An empty string can be used in a rule. It means that the variable is absent from the rule. It will not be taken into account in the inference process.

The Active column allows to activate/inactivate rules.

Warning: Only active rules are stored when saving the FIS configuration file.

Rules are displayed as a table (1 column per input/output). They can be sorted by column.

An option in the Display menu allows to display or not display inactive variables. This option is useful to deal with highly dimensional systems.

Another option in the Display menu allows to give a weight to rules. The weights must be positive and are stored in the FIS configuration file, with 3 decimal digits, as soon as one weight is given a value different from default, which is equal to 1. Expert weights change the inference result. All rule matching degrees are multiplied by the corresponding expert weights at the rule aggregation functions (sum or max). Note: for fuzzy outputs, the effect of weights greater than 1 may may saturate quickly. For fuzzy output defuzzification operators, except SugenoFuzzy, the weight sets the alpha-cut level of the relevant output MF. It is therefore lower than 1.