evalfis

Evaluate fuzzy inference system

Syntax

output = evalfis(fis,input)

output = evalfis(fis,input,options)

[output,fuzzifiedIn,ruleOut,aggregatedOut,ruleFiring] = evalfis(___)

Description

output = evalfis(fis,input) evaluates the fuzzy inference system fis for the input values in input and returns the resulting output values.

output = evalfis(fis,input,options) evaluates the fuzzy inference system using specified evaluation options.

[output,fuzzifiedIn,ruleOut,aggregatedOut,ruleFiring] = evalfis(___) returns intermediate results from the fuzzy inference process. This syntax is not supported when fis is a fistree object.

Examples

Evaluate Fuzzy Inference System

Load FIS.

fis = readfis("tipper");

Evaluate the FIS when the first input is 2 and the second input is 1.

output = evalfis(fis,[2 1])

output = 
7.0169

Evaluate FIS for Multiple Input Combinations

Load FIS.

fis = readfis("tipper");

Specify the input combinations to evaluate using an array with one row per input combination.

input = [2 1;
         4 5;
         7 8];

Evaluate the FIS for the specified input combinations.

output = evalfis(fis,input)

output = 3×1

    7.0169
   14.4585
   20.3414

Each row of output is the defuzzified output value for the corresponding row of input.

Specify Number of Output Samples for FIS Evaluation

Load FIS.

fis = readfis("tipper");

Create an evalfisOptions option set, specifying the number of samples in the output fuzzy sets.

options = evalfisOptions(NumSamplePoints=50);

Evaluate the FIS using this option set.

output = evalfis(fis,[2 1],options);

Evaluate Tree of Fuzzy Inference Systems

Create a pair of Mamdani fuzzy inference systems.

fis1 = mamfis(Name="fis1",NumInputs=2,NumOutputs=1);
fis2 = mamfis(Name="fis2",NumInputs=2,NumOutputs=1);

Define the connection between the two.

con = ["fis1/output1" "fis2/input1"];

Create a tree of fuzzy inference systems.

tree = fistree([fis1 fis2],con);

Create an evalfisOptions option set, specifying the number of samples in the output fuzzy sets.

options = evalfisOptions(NumSamplePoints=50);

Evaluate the fistree object using a specified input combination and this option set.

y = evalfis(tree,[0.5 0.2 0.7],options)

y = 
0.1553

Obtain Intermediate Fuzzy Inference Results

Load FIS.

fis = readfis("tipper");

Evaluate the FIS, and return the intermediate inference results.

[output,fuzzifiedIn,ruleOut,aggregatedOut,ruleFiring] = evalfis(fis,[2 1]);

You can examine the intermediate results to understand or visualize the fuzzy inference process. For example, view the aggregated output fuzzy set, which is the fuzzy set that evalfis defuzzifies to find the output value. Also, plot the defuzzified output value.

outputRange = linspace(fis.output.range(1),fis.output.range(2),length(aggregatedOut))'; 
plot(outputRange,aggregatedOut,[output output],[0 1])
xlabel("Tip")
ylabel("Output Membership")
legend("Aggregated output fuzzy set","Defuzzified output")

The length of aggregatedOutput corresponds to the number of sample points used to discretize output fuzzy sets.

Evaluate Type-2 Fuzzy Inference System

Create a type-2 Mamdani fuzzy inference system.

fis = mamfistype2(NumInputs=2,NumOutputs=1);

Evaluate the FIS when the first input is 0.4 and the second input is 0.72.

output = evalfis(fis,[0.4 0.72])

output = 
0.1509

The output of a type-2 FIS is a crisp value.

When you obtain intermediate fuzzy inference results for a type-2 FIS, you obtain intermediate results generated using both upper and lower MF values. For example, obtain the intermediate fuzzified input values.

[output,fuzzifiedInput] = evalfis(fis,[0.5 0.75]);

View the fuzzified input values.

fuzzifiedInput

fuzzifiedInput = 9×4

         0         0         0         0
    1.0000         0    1.0000         0
         0         0         0         0
         0    0.4000         0    0.2500
    1.0000    0.4000    1.0000    0.2500
         0    0.4000         0    0.2500
         0    0.4000         0    0.2500
    1.0000    0.4000    1.0000    0.2500
         0    0.4000         0    0.2500

The first two columns contain the fuzzified values of the first and second inputs based on the upper MF for each input. The second two columns contain the fuzzified values for based on the lower MF for each input.

Input Arguments

fis — Fuzzy inference system

mamfis object | sugfis object | mamfistype2 object | sugfistype2 object | fistree object | homogeneous structure

Fuzzy inference system (FIS) to be evaluated, specified as one of the following:

• mamfis object — Mamdani FIS

• sugfis object — Sugeno FIS

• mamfistype2 object — Type-2 Mamdani FIS

• sugfistype2 object — Type-2 Sugeno FIS

• fistree object — Tree of interconnected FISs

• Homogeneous structure created using getFISCodeGenerationData. For an example, see Generate Code for Fuzzy System Using MATLAB Coder.

input — Input values

M-by-N_U array

Input values, specified as an M-by-N_U array, where N_U is the number of input variables in fis and M is the number of input combinations to evaluate.

evalfis supports double-precision or single-precision input values.

options — Evaluation options

evalfisOptions object

Evaluation options, specified as an evalfisOptions object.

Output Arguments

output — Output values

array

Output values, returned as an M-by-N_Y array, where N_Y is the number of output variables in fis. evalfis evaluates fis for each row of input and returns the resulting defuzzified outputs in the corresponding row of output.

fuzzifiedIn — Fuzzified input values

array

Fuzzified input values, returned as an array.

When fis is a type-1 fuzzy inference system, fuzzifiedIn is an N_R-by-N_U array, where N_R is the number of rules in fis. Element (i,j) of fuzzifiedIn is the value of the input membership function for the jth input in the ith rule.

When fis is a type-2 fuzzy inference system, fuzzifiedIn is an N_R-by-(2*N_U) array. The first N_U columns contain the fuzzified values of the upper membership function for each rule, and the last N_U columns contain the fuzzified values from the lower membership functions.

If input specifies multiple input combinations, then fuzzifiedIn corresponds to the combination in the last row of input.

For more information on fuzzifying input values, see Fuzzify Inputs.

This output argument is not supported when fis is a fistree object.

ruleOut — Rule outputs

array

Rule outputs, returned as an array. To obtain the output for each rule, evalfis applies the firing strength from the rule antecedent to the output membership function using the implication method specified in fis.

When fis is a type-1 Mamdani system, ruleOut is an N_S-by-(N_RN_Y) array, where N_R is the number of rules, N_Y is the number of outputs, and N_S is the number of sample points used for evaluating output variable ranges. Each column of ruleOut contains the output fuzzy set for one rule. The first N_R columns contain the rule outputs for the first output variable, the next N_R columns correspond to the second output variable, and so on.

When fis is a type-2 Mamdani system, ruleOut is an N_S-by-(2*N_R*N_Y) array. The first N_R*N_Y columns contain the rule outputs generated using upper membership functions, and the last N_R*N_Y columns contain the rule outputs generated using lower membership functions.

When fis is a type-1 Sugeno system, each rule output is a scalar value. In this case, ruleOut is an N_R-by-N_Y array. Element (j,k) of ruleOut is the value of the kth output variable for the jth rule.

When fis is a type-2 Sugeno system, ruleOut is an N_R-by-(3*N_Y) array. The first N_Y columns contain the rule output levels. The next N_Y columns contain the corresponding rule firing strengths generated using upper membership functions. The last N_Y columns contain the rule firing strengths generated using lower membership functions. For example, in a three-output system, columns 4 and 7 contain the firing strengths for the output levels in column 1.

If input specifies multiple input combinations, then ruleOut corresponds to the combination in the last row of input.

For more information on fuzzy implication, see Apply Implication Method.

This output argument is not supported when fis is a fistree object.

aggregatedOut — Aggregated output

array | row vector

Aggregated output for each output variable, returned as an N_S-by-N_Y array or a row vector of length N_Y. For each output variable, evalfis combines the corresponding outputs from all the rules using the aggregation method specified in fis.

For a type-1 Mamdani system, the aggregate result for each output variable is a fuzzy set. In this case, aggregatedOut is as an N_S-by-N_Y array, where N_Y is the number of outputs and N_S is the number of sample points used for evaluating output variable ranges. Each column of aggregatedOut contains the aggregate fuzzy set for one output variable.

For a type-2 Mamdani system, the aggregate result for each output variable is a fuzzy set. In this case, aggregatedOut is as an N_S-by-(2*N_Y) array. The first N_Y columns contain the aggregated outputs generated using upper membership functions, and the last N_Y columns contain the aggregated outputs generated using lower membership functions.

When fis is a type-1 Sugeno system, the aggregate result for each output variable is a scalar value. In this case, aggregatedOut is a row vector of length N_Y, where element k is the sum of the rule outputs for the kth output variable.

When fis is a type-2 Sugeno system, aggregatedOut is an N_R-by-(3*N_Y) array. aggregatedOut contains the same data as ruleOut with the columns sorted based on the output levels. For example, in a three-output system, when the output levels in column 1 are sorted, the corresponding firing strengths in columns 4 and 7 are adjusted accordingly.

If input specifies multiple input combinations, then aggregatedOut corresponds to the combination in the last row of input.

For more information on fuzzy aggregation, see Aggregate All Outputs.

This output argument is not supported when fis is a fistree object.

ruleFiring — Rule firing strengths

column vector | array

Rule firing strength, returned as a column vector or array. To obtain the firing strength for each rule, evalfis evaluates the rule antecedents; that is, it applies fuzzy operator to the values of the fuzzified inputs.

For a type-1 fuzzy system, ruleFiring is a column vector of length N_R, where N_R is the number of rules, and element i is the firing strength of the ith rule.

For a type-2 fuzzy system, ruleFiring is an N_R-by-2 array. The first column contains the rule firing strengths generated using upper membership functions, and the second column contains the rule firing strengths generated using lower membership functions.

If input specifies multiple input combinations, then ruleFiring corresponds to the combination in the last row of input.

For more information on applying the fuzzy operator, see Apply Fuzzy Operator.

This output argument is not supported when fis is a fistree object.

Alternative Functionality

App

You can interactively evaluate type-1 and type-2 fuzzy inference systems using the Fuzzy Logic Designer app. The app does not support FIS trees.

Simulink Block

You can evaluate FIS objects using the Fuzzy Logic Controller block. This block does not support evaluating FIS trees. For more information on mapping the arguments of evalfis to the Fuzzy Logic Controller block, see Simulate Fuzzy Inference Systems in Simulink.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• All evalfis syntaxes are supported for code generation. However, mamfis, sugfis, mamfistype2, sugfistype2, and fistree objects are not supported. To use evalfis for code generation, you must convert your FIS objects into homogeneous structures using getFISCodeGenerationData.

• Unlike the Fuzzy Logic Controller block, evalfis does not support fixed-point data for simulation or code generation.

• When evaluating a fuzzy inference system in Simulink^®, it is recommended to not use evalfis or evalfisOptions within a MATLAB Function block. Instead, evaluate your fuzzy inference system using a Fuzzy Logic Controller block.

Version History

Introduced before R2006a

R2024b: Fuzzy inference system structures not supported

evalfis no longer supports fuzzy inference system structures. Use mamfis and sugfis objects instead. To convert existing fuzzy inference system structures to objects, use the convertfis function.

R2022b: C and C++ code generation support for evaluating FIS trees

You can generate code for evaluating FIS trees using MATLAB^® Coder™. For more information, see Generate Code for Fuzzy System Using MATLAB Coder.

R2019b: Fuzzy inference system structures will not be supported

Support for fuzzy inference systems structures will be removed in a future release. This change was announced in R2018b. Using fuzzy inference system structures with evalfis issues a warning starting in R2019b.

R2018b: C and C++ code generation support

You can generate code for loading and evaluating the following types of fuzzy inference systems using MATLAB Coder.

• Type-1 and type-2 Mamdani systems

• Type-1 and type-2 Mamdani systems

For more information, see Generate Code for Fuzzy System Using MATLAB Coder.

R2018b: Input argument order changed

Previously, to evaluate a fuzzy inference system, fis, you specified the input variable values, input, as the first input argument. For example:

output = evalfis(input,fis);
output = evalfis(input,fis,options);

Update your code to specify the fuzzy inference system as the first input argument. For example:

output = evalfis(fis,input);
output = evalfis(fis,input,options);

R2018b: Obtain intermediate rule firing strengths

You can obtain intermediate rule firing strengths when evaluating a fuzzy inference system using the ruleFiring output argument.

R2018a: Specify number of sample points for output fuzzy sets using evalfisOptions object

Previously, to specify the number of sample points, numPts, to use when evaluating output fuzzy sets of fuzzy inference system fis, you used an input argument. For example:

output = evalfis(input,fis,numPts);

Update your code to specify the number of sample points using an evalfisOptions object. For example:

opt = evalfisOptions('NumSamplePoints',numPts);
output = evalfis(input,fis,opt);

R2018a: Diagnostic message behavior changed

The diagnostic message behavior of the evalfis function has changed. Previously, the evalfis function had the following behaviors for diagnostic conditions.

Diagnostic Condition	Previous Behavior
Input values outside of their specified variable ranges	MATLAB warning
No rules fired for a given output at the current input values	MATLAB Command Window message
Empty output fuzzy sets	MATLAB Command Window message

These diagnostic conditions are reported as MATLAB warnings by default. You can change this behavior by specifying the corresponding options in an evalfisOptions object.

To disable the default warning messages, update your code to use an evalfisOptions object, and specify the diagnostic message options. For example, disable the empty output fuzzy set message.

opt = evalfisOptions('EmptyOutputFuzzySetMessage',"none");
output = evalfis(input,fis,opt);

R2018a: Intermediate fuzzy inference outputs for Sugeno systems analogous to outputs for Mamdani systems

When evaluating a Sugeno system using the following syntax, the intermediate fuzzy inference results are now analogous to the intermediate results for Mamdani systems.

[output,fuzzifiedInputs,ruleOutputs,aggregatedOutput] = evalfis(input,fis);

For a Sugeno system:

• ruleOutputs now returns an array that contains the scalar output value for each rule; that is, the product of the rule firing strength and the rule output level.

• aggregatedOutput now returns the sum of all the rule output values for each output variable.

Previously, for a Sugeno fuzzy system:

• ruleOutputs returned an array that contained the output level for each rule.

• aggregatedOutput returned an array that contained the firing strength for each rule.

Starting in R2018a, if your code returns intermediate fuzzy inference results when evaluating a Sugeno system using evalfis, modify your code to use the new ruleOutputs and aggregatedOutput results.

R2017b: Smaller number of sample points supported for output variable ranges

Using the numPts input argument, you can specify the number of sample points for evaluating the output range of a Mamdani fuzzy inference system as any value greater than 1. Previously, the minimum value was 101.