opt = anfisOptions creates
a default option set for tuning a Sugeno fuzzy inference system using anfis. Use dot notation to modify this option
set for your specific application. Any options that you do not modify
retain their default values.
Specify optional
comma-separated pairs of Name,Value arguments. Name is
the argument name and Value is the corresponding value.
Name must appear inside quotes. You can specify several name and value
pair arguments in any order as
Name1,Value1,...,NameN,ValueN.
Example: 'EpochNumber',50 sets the maximum number of training epochs to
50.
Initial FIS structure to tune, specified as the comma-separated
pair consisting of 'InitialFIS' and one of the
following:
Positive integer greater than 1 specifying
the number of membership functions for all input variables. anfis generates
an initial FIS structure with the specified number of membership functions
using genfis with grid partitioning.
Vector of positive integers with length equal to the
number of input variables specifying the number of membership functions
for each input variable. anfis generates an initial
FIS structure with the specified numbers of membership functions using genfis with
grid partitioning.
FIS structure generated using genfis command
with grid partitioning or subtractive clustering. The specified system
must have the following properties:
Single output, obtained using weighted average defuzzification.
First or zeroth order Sugeno-type system; that is,
all output membership functions must be the same type and be either 'linear' or 'constant'.
No rule sharing. Different rules cannot use the same
output membership function; that is, the number of output membership
functions must equal the number of rules.
Unity weight for each rule.
No custom membership functions or defuzzification
methods.
'EpochNumber' — Maximum number of training epochs 10 (default) | positive integer
Maximum number of training epochs, specified as the comma-separated
pair consisting of 'EpochNumber' and a positive
integer. The training process stops when it reaches the maximum number
of training epochs.
'ErrorGoal' — Training error goal 0 (default) | scalar
Training error goal, specified as the comma-separated pair consisting
of 'ErrorGoal' and a scalar. The training process
stops when the training error is less than or equal to ErrorGoal.
Initial training step size, specified as the comma-separated
pair consisting of 'InitialStepSize' and a positive
scalar.
The anfis training algorithm tunes the
FIS parameters using gradient descent optimization methods. The training
step size is the magnitude of each gradient transition in the parameter
space. Typically, you can increase the rate of convergence of the
training algorithm by increasing the step size. During optimization, anfis automatically
updates the step size using StepSizeIncreaseRate and StepSizeDecreaseRate.
Generally, the step-size profile during
training is a curve that increases initially,
reaches some maximum, and then decreases for the
remainder of the training. To achieve this ideal
step-size profile, adjust the initial step-size
and the increase and decrease rates
(opt.StepSizeDecreaseRate,
opt.StepSizeIncreaseRate).
'StepSizeDecreaseRate' — Step-size decrease rate 0.9 (default) | positive scalar less than
1
Step-size decrease rate, specified as the
comma-separated pair consisting of
'StepSizeDecreaseRate' and a
positive scalar less than 1. If
the training error undergoes two consecutive
combinations of an increase followed by a
decrease, then anfis scales
the step size by the decrease rate.
Step-size increase rate, specified as the
comma-separated pair consisting of
'StepSizeIncreaseRate' and a
scalar greater than 1. If the
training error decreases for four consecutive
epochs, then anfis scales the
step size by the increase rate.
'DisplayANFISInformation' — Flag for showing ANFIS information 1 (default) | 0
Flag for showing ANFIS information at the start of the training
process, specified as the comma-separated pair consisting of 'DisplayANFISInformation' and
one of the following:
1 — Display the following
information about the ANFIS system and training
data:
Number of nodes in the ANFIS system
Number of linear parameters to tune
Number of nonlinear parameters to
tune
Total number of parameters to tune
Number of training data pairs
Number of checking data pairs
Number of fuzzy rules
0 — Do not display the
information.
'DisplayErrorValues' — Flag for showing training error values 1 (default) | 0
Flag for showing training error values after each training epoch,
specified as the comma-separated pair consisting of 'DisplayErrorValues' and
one of the following:
1 — Display the training
error.
0 — Do not display the
training error.
'DisplayStepSize' — Flag for showing step size 1 (default) | 0
Flag for showing step size whenever the step size changes, specified
as the comma-separated pair consisting of 'DisplayStepSize' and
one of the following:
1 — Display the step
size.
0 — Do not display the
step size.
'DisplayFinalResults' — Flag for displaying final results 1 (default) | 0
Flag for displaying final results after training, specified
as the comma-separated pair consisting of 'DisplayFinalResults' and
one of the following:
1 — Display the
results.
0 — Do not display the
results.
'ValidationData' — Validation data [] (default) | array
Validation data for preventing overfitting to the training data,
specified as the comma-separated pair consisting of 'ValidationData' and
an array. For a fuzzy system with N inputs, specify ValidationData as
an array with N+1 columns. The first N columns
contain input data and the final column contains output data. Each
row of ValidationData contains one data point.
At each training epoch, the training algorithm validates the
FIS using the validation data.
Generally, validation data should fully represent the features
of the data the FIS is intended to model, while also being sufficiently
different from the training data to test training generalization.
Optimization method used in membership function parameter training,
specified as the comma-separated pair consisting of 'OptimizationMethod' and
one of the following:
1 — Use a hybrid method,
which uses a combination of backpropagation to
compute input membership function parameters, and
least squares estimation to compute output
membership function parameters.
0 — Use backpropagation
gradient descent to compute all parameters.
You can also select a web site from the following list:
How to Get Best Site Performance
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.