Fuzzy system tuning with ACP + relative distances as parameters
fis=SystemTuning2(fis,InTrain,OutTrain,params,options,etTuningInfo,axPIDiagram,axOutputDataDiagram)
Fuzzy system tuning using a "hill climbing" approach.
It works with MISO systems and trapezoid shaped
linguistic terms ('trapmf'). Using the method ACP.
Parameterization mode: relative distances as parameters.
-------------------------------------------------------------------------
Parameters
-------------------------------------------------------------------------
fis - complex structure describing the fuzzy inference system
InTrain - N by M matrix containing the input data, where N is number of
data points and M is the number of input dimensions
OutTrain - N by 1 matrix containing the output data, where N is number of
data points and 1 is the number of output dimensions
params - parameters of the fuzzy inference
options - Options of the system tuning
options(1) Number of allowed decimals
options(2) Coefficient for the step
options(3) SimilarityTreshold
Indicates the treshold value below which two sets
are merged in one. The treshold is expressed in
proportion of the range of the actual input
variable. Its value is inside the interval [0,1].
options(4) Number of iterations
options(5) info display during iteration
options(6) Save FIS and info after each calculation step
options(7) IncrementTreshold
Increase (double) step if the amelioration of the
performance index is greater than
IncrementTreshold during an iteration.
options(8) DecrementTreshold
Decrease (halve) step if the amelioration of the
performance index is smaller than
DecrementTreshold during an iteration.
options(9) Merge simlar sets after each iteration
options(10) Merge similar rules after each iteration
options(11) Plot Performance Index
Options(12) Plot original and calculated data
options(13) Tuning method type
options(14) Performance index type
options(15) Stop the iteration when the performance index
falls below this value. It is expressed in proportion of
the range of the output linguistic variable.
options(16) Core width in proportion of the range
options(17) Support width in proportion of the range
etTuningInfo - handle of the editbox displaying the tuning preferences
axPIDiagram - handles of the axes PI
axOutputDataDiagram - handles of the OutputDataDiagram
-------------------------------------------------------------------------
Returned values
-------------------------------------------------------------------------
fis - complex structure describing the fuzzy inference system
-------------------------------------------------------------------------
Remarks
-------------------------------------------------------------------------
The centre of the core is is used as reference point.
-------------------------------------------------------------------------
Zsolt Csaba Johanyák, johanyak.csaba@gamf.kefo.hu, v. 4.1, 30 Sep 2007.