How to resolve the error: the variable fuzout(1,i) cannot save fuzout = evalfis(fis, trainin)

Hello, i'm getting an error in my code:

Unable to perform assignment because the indices on the left side are not compatible with

the size of the right side.

Error in anfisloop (line 6)

[fuzout(1,i),trnRMSE(1,i)] = evaltrainset(T(1,i),trainin,trainou);

my code: the function i'm calling:

this function will create 5 models each one will be in a column in T[].

so T(1,1) is the first model, T(1,2) the second .....

the idea of what i want to do is:

create a variable (here fuzout[]) that will stack in each column the output of a model

and another variable that will stack in each column the RMSE of a model.

function [T,fuzout,trnRMSE] = anfisloop(trainin,trainou,testin,testout,fis)

epoch=5;

for i = 1:epoch

opt=anfisopt(testin,testout,fis,i);

T(1,i) = anfis([trainin trainou],opt);

[fuzout(1,i),trnRMSE(1,i)] = evaltrainset(T(1,i),trainin,trainou);

end

end

the function evaltrainset that the precious function call:

evalfis is a matlab's function

function [fuzout,trnRMSE] = evaltrainset(fis,trainin,trainout)

fuzout = evalfis(fis,trainin);

trnRMSE = norm(fuzout-trainout)/sqrt(length(fuzout));

end

how i am calling the function :

>> [A,fuzout,trnRMSE]=anfisloop(trainin,trainout,testin,testout,fis)

please help.

[Torsten]

I doubt that the output from anfis can be saved as a scalar:

T(1,i) = anfis([trainin trainou],opt);

Try

T{i} = anfis([trainin trainou],opt);
[fuzout(1,i),trnRMSE(1,i)] = evaltrainset(T{i},trainin,trainou);

[redha wassim brahimi]

T(1,i) = anfis([trainin trainou],opt);

works normally it saves all the models .

the problem was in

[fuzout(1,i),trnRMSE(1,i)] = evaltrainset(T{i},trainin,trainou);

the variable fuzout(1,i) can not save

fuzout = evalfis(fis,trainin);

from the other function and i think it is because : the output of evalfis is a table and i want to save a table in fuzout(1,i) .

[Sam Chak]

Hey @redha wassim brahimi

@Torsten is right. The output of anfis(trainingData, options) is a FIS object. I guess you want to create one FIS in each loop and then store it?

[redha wassim brahimi]

hey @Sam Chak

yes that's what i want to do , is my code wrong in this ?

Answer 1

[Sam Chak]

Hi @redha wassim brahimi

I am revisiting the problem after nearly three years. I created dummy data and resolved the issue involving the evaltrainset() command in the anfisloop() function. The reason you are unable to perform the assignment in that line is that the right side, evalfis(fis, trainin), returns a vector, while the left side, fuzout(1,i), is only capable of storing a single value at each loop. The correct syntax for the left side should be fuzout(:,i).

%% custom function "anfisloop" to create one FIS in each loop
function [T, fuzout, trnRMSE] = anfisloop(trainin, trainou, testin, testout, fis, epoch)
   for i = 1:epoch
        opt     = anfisOptions('InitialFIS', fis, 'EpochNumber', i, 'ValidationData', [testin, testout]);
        opt.DisplayANFISInformation = 0;
        opt.DisplayErrorValues      = 0;
        opt.DisplayStepSize         = 0;
        opt.DisplayFinalResults     = 0;
        T(1,i)  = anfis([trainin trainou], opt);
        [fuzout(:,i), trnRMSE(1,i)] = evaltrainset(T(1,i), trainin, trainou);
   end
end

%% custom function "evaltrainset" to evaluate FIS
function [fuzout, trnRMSE] = evaltrainset(fis, trainin, trainou)
   fuzout   = evalfis(fis, trainin);
   trnRMSE  = norm(fuzout - trainou)/sqrt(length(fuzout));
end

%% dummy data
testin  = linspace(-1, 1, 201)';
testout = 2/pi*atan(5*pi/2*testin);
trainin = testin;
trainou = erf(5*trainin);

%% generate initial FIS
fis = genfis(testin, testout);

%% train ANFIS model
epoch   = 5;
[T, fuzout, trnRMSE] = anfisloop(trainin, trainou, testin, testout, fis, epoch);

%% display predictions of ANFIS models
figure
plot(trainin, fuzout), grid on
title('Plots of fuzout')

●

%% display prediction accuracy of each ANFIS model
disp(trnRMSE)

    0.0986    0.0976    0.0965    0.0954    0.0944