How to resolve the issue of Y must be a column vector?

template = templateTree('MaxNumSplits', 20);
IncMdl = fitcensemble(Xinit, Yinit, 'Method', 'Bag', 'Learners', template);

Unrecognized function or variable 'Xinit'.

Given below code is not updating the model. It gives error that Yinc must be a vector. I have checked that Yinc is a vector. Please suggest me how to resolve this issue

code is here

startIdx = incrementSize + 1;
while startIdx <= length(trainLabels)
    endIdx = min(startIdx + incrementSize - 1,length(trainLabels));
    Xinc = trainFeatures(startIdx:endIdx,:);
    Yinc = trainLabels(startIdx:endIdx);
    % Ensure Yinc is a column vector and categorical
    Yinc = Yinc(:);
    if ~iscategorical(Yinc)
        Yinc = categorical(Yinc);
    end
    % Update model incrementally
    IncMdl = fit(IncMdl, Xinc, Yinc);
    startIdx = endIdx + 1;
end

[Cris LaPierre]

Please share the full error message (all the red text). We can also be more helpful if you can share the data needed to run the code. Consider saving your variables to a mat file and attaching that to your post using the paperclip icon.

[Aksh Kumar]

The complete error message is given below

Error using fit>iFit (line 140)

Y must be a column vector.

Error in fit (line 117)

[fitobj, goodness, output, convmsg] = iFit( xdatain, ydatain, fittypeobj, ...

Error in SAR_Classification (line 178)

IncMdl = fit(IncMdl, Xinc, Yinc);

The size of data file is 570mb, Please suggest me how to upload it.

Thanks for your kind help.

[Cris LaPierre]

Here is a minimal example that reproduces your error.

load ionosphere

Xinit = X;
Yinit = Y;
trainLabels = unique(Yinit);
trainFeatures = X;

template = templateTree('MaxNumSplits', 20);
IncMdl = fitcensemble(Xinit, Yinit, 'Method', 'Bag', 'Learners', template);

incrementSize = 1;
startIdx = incrementSize + 1;
while startIdx <= length(trainLabels)
    endIdx = min(startIdx + incrementSize - 1,length(trainLabels));
    Xinc = trainFeatures(startIdx:endIdx,:);
    Yinc = trainLabels(startIdx:endIdx);
    % Ensure Yinc is a column vector and categorical
    Yinc = Yinc(:);
    if ~iscategorical(Yinc)
        Yinc = categorical(Yinc);
    end
    % Update model incrementally
    IncMdl = fit(IncMdl, Xinc, Yinc);
    startIdx = endIdx + 1;
end

Error using fit>iFit (line 140)
Y must be a column vector.

Error in fit (line 117)
[fitobj, goodness, output, convmsg] = iFit( xdatain, ydatain, fittypeobj, ...

When I run which fit I get the following

C:\Program Files\MATLAB\R2024b\toolbox\curvefit\curvefit\fit.m

confirming Walter's suspicion that it is calling fit in the Curve Fitting toolbox.

Answer 1

[Walter Roberson]

IncMdl = fitcensemble(Xinit, Yinit, 'Method', 'Bag', 'Learners', template);

The result of fitting with method 'bag' is a ClassificationBaggedEnsemble https://www.mathworks.com/help/stats/classreg.learning.classif.classificationbaggedensemble.html

ClassificationBaggedEnsemble do not have any property or object functions named fit so the call

IncMdl = fit(IncMdl, Xinc, Yinc);

is probably being intepreted as a call to fit from the Curve Fitting Toolbox.

Note that there are various models that offer fit() function, including incrementalClassificationLinear ... but bagged ensemble does not offer fit(). None of the models potentially generated by fitcensemble() offer fit()

Once more you are trying to do incremental learning on a fitcensemble object, which is something that cannot be done.

[Aksh Kumar]

Thanks for your suggestion. I am trying to use it

IncMdl = incrementalClassificationLinear(Xinit, Yinit); 

however it is giving following error

Error using internal.stats.parseArgs (line 43)

Parameter name must be text.

Error in incremental.utils.parseUnsupportedConstructionParameters>parseUnsupportedConstructionParametersHelper (line 42)

[~,userInput,varargin] = internal.stats.parseArgs(param,[],varargin{:});

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in incremental.utils.parseUnsupportedConstructionParameters (line 25)

varargin = parseUnsupportedConstructionParametersHelper(nvparams,varargin{:});

Error in incrementalClassificationLinear (line 392)

varargin = incremental.utils.parseUnsupportedConstructionParameters({'MaxNumClasses'},{'MaxNumClasses'}, varargin{:});

Error in SAR_Classification (line 150)

IncMdl = incrementalClassificationLinear(Xinit, Yinit);

please suggest me how to define it.

[Walter Roberson]

incrementalClassificationLinear() accepts only name/value pairs. None of the named options have anything to do with X or Y values.

The flow of control would be more like

IncMdl = incrementalClassificationLinear(); 
IncMdl = fit(IncMdl, Xinc, Yinc);

[Aksh Kumar]

it is giving following error

Error using incremental.classif.ClassificationModel/addClasses (line 707)

Incremental model must have 2 unique class labels.

but in my case there are 6 unique class labels.

please suggest me incremental classification model who can take multiple unique class labels.

[Aksh Kumar]

i am trying to use the following model for multiple class labels

IncMdl = incrementalClassificationECOC(); 
IncMdl = fit(IncMdl, Xinit, Yinit,'MaxNumClasses', 6);

however, it is giving following error

Error using incremental.classif.ClassificationModel/setMaxNumClasses (line 246)

'MaxNumClasses' or 'ClassNames' must be provided.

please suggest me how use it.

[Walter Roberson]

IncMdl = incrementalClassificationECOC(MaxNumClasses=6); 
IncMdl = fit(IncMdl, Xinit, Yinit);

[Cris LaPierre]

Reusing the example from above, this incorporates Walter's suggestion.

load ionosphere

Xinit = X;
Yinit = Y;
trainLabels = unique(Yinit);
trainFeatures = X;

template = templateTree('MaxNumSplits', 20);
IncMdl = fitcensemble(Xinit, Yinit, 'Method', 'Bag', 'Learners', template);

incrementSize = 1;
startIdx = incrementSize + 1;
while startIdx <= length(trainLabels)
    endIdx = min(startIdx + incrementSize - 1,length(trainLabels));
    Xinc = trainFeatures(startIdx:endIdx,:);
    Yinc = trainLabels(startIdx:endIdx);
    % Ensure Yinc is a column vector and categorical
    Yinc = Yinc(:);
    if ~iscategorical(Yinc)
        Yinc = categorical(Yinc);
    end
    % Update model incrementally
    IncMdl = incrementalClassificationLinear(); 
    IncMdl = fit(IncMdl, Xinc, Yinc);
    startIdx = endIdx + 1;
end

[Walter Roberson]

Probably the

IncMdl = incrementalClassificationLinear(); 

should be before the while loop.

[Aksh Kumar]

Thanks. I tried to use this function given below as

IncMdl = incrementalClassificationLinear();

IncMdl = fit(IncMdl, Xinc, Yinc);

error is

Error using incremental.classif.ClassificationModel/addClasses (line 707)

Incremental model must have 2 unique class labels.

I request you to please suggest me incrementalClassification model for multiple classes.

[Walter Roberson]

https://www.mathworks.com/help/stats/incrementalclassificationlinear.html

"incrementalClassificationLinear creates an incrementalClassificationLinear model object, which represents a binary classification linear model for incremental learning."

So incrementalClassificationLinear is inherently two-class.

"For multiclass incremental learning, see incrementalClassificationECOC and incrementalClassificationNaiveBayes."