Undefined function 'round' for input arguments of type 'categorical' error

Hello experts,

I was trying to perform the example of function fitclinear described in the matlab help documentation, and I got an error.

Here is the described the example:

load nlpdata

X = X';
Ystats = Y == 'stats';

rng default
Mdl = fitclinear(X,Ystats,'ObservationsIn','columns','Solver','sparsa',...
    'OptimizeHyperparameters','auto','HyperparameterOptimizationOptions',...
    struct('AcquisitionFunctionName','expected-improvement-plus'))

Here the error that I got:

Undefined function 'round' for input arguments of type 'categorical'.

Error in iscategorical (line 26)
if size(mat, 1) == 1 && all(round(mat) == mat)

Error in classreg.learning.paramoptim.prepareArgValue (line 12)
elseif iscategorical(elt)

Error in classreg.learning.paramoptim.BayesoptInfo.argsFromTable (line 151)
                        classreg.learning.paramoptim.prepareArgValue(XTable{1,v})}];

Error in classreg.learning.paramoptim.BayesoptInfo/updateArgsFromTable (line 56)
            ArgsFromTable = classreg.learning.paramoptim.BayesoptInfo.argsFromTable(XTable);

Error in classreg.learning.paramoptim.createObjFcn/theObjFcn (line 17)
        NewFitFunctionArgs = updateArgsFromTable(BOInfo, FitFunctionArgs, XTable);

Error in BayesianOptimization/callObjNormally (line 2184)
                        Objective = this.ObjectiveFcn(conditionalizeX(this, X));

Error in BayesianOptimization/callObjFcn (line 2145)
                    = callObjNormally(this, X);

Error in BayesianOptimization/performFcnEval (line 2128)
                ObjectiveFcnObjectiveEvaluationTime, this] = callObjFcn(this, this.XNext);

Error in BayesianOptimization/run (line 1836)
                this = performFcnEval(this);

Error in BayesianOptimization (line 450)
            this = run(this);

Error in bayesopt (line 287)
Results = BayesianOptimization(Options);

Error in classreg.learning.paramoptim.fitoptimizing>doBayesianOptimization (line 182)
OptimizationResults = bayesopt(objFcn, VariableDescriptions, ...

Error in classreg.learning.paramoptim.fitoptimizing (line 136)
        [OptimizationResults, XBest] = doBayesianOptimization(objFcn, BOInfo, VariableDescriptions,
        HyperparameterOptimizationOptions);

Error in fitclinear (line 218)
    [varargout{1:nargout}] = classreg.learning.paramoptim.fitoptimizing('fitclinear',X,y,varargin{:});

Any suggestions will be helpful.

Thanks

[dpb]

What does

whos X

return? categorical class, I presume?

Try

Mdl = fitclinear(double(X),Ystats,'ObservationsIn','columns','Solver','sparsa',...
    ...

instead and see if joy ensues. It surely is an irritation when examples don't work either because they simply were pasted in from non-working code or TMW has made changes in other routines that may have been allowable in earlier versions and no longer are.

[Walter Roberson]

I did not encounter this problem when I tested in R2019a on my Mac.

[Walter Roberson]

What shows up for

which -all iscategorical

The main iscategorical() routine is quite short. The one for classreg.learning.internal.ClassLabel does not have any round() in it, at least in R2019a.

[dpb]

With R2017b it errors...my R2018b install got blown away by a Window update and haven't had the time to reinstall it...so, it's probably from OP looking at new doc and running older release.

[abdulkawi alareefi]

My matlab version is R2017a.

>> whos X
  Name          Size                  Bytes  Class     Attributes

  X         34023x31572            36696696  double    sparse    

>> which -all iscategorical
G:\1master courses\semester 2\2015_Advances in brain imaging\software\princeton-mvpa-toolbox-master\core\util\iscategorical.m
D:\Program Files\MATLAB\R2017a\toolbox\distcomp\parallel\@codistributed\iscategorical.m                                        % codistributed method
D:\Program Files\MATLAB\R2017a\toolbox\stats\classreg\+classreg\+learning\+internal\ClassLabel.m                               % classreg.learning.internal.ClassLabel method
D:\Program Files\MATLAB\R2017a\toolbox\matlab\datatypes\iscategorical.m                                                        % Shadowed

I also tried double(X)

>> Mdl = fitclinear(double(X),Ystats,'ObservationsIn','columns','Solver','sparsa',...
    'OptimizeHyperparameters','auto','HyperparameterOptimizationOptions',...
    struct('AcquisitionFunctionName','expected-improvement-plus'))
Undefined function 'round' for input arguments of type 'categorical'.

Error in iscategorical (line 26)
if size(mat, 1) == 1 && all(round(mat) == mat)

Answer 1

[Walter Roberson]

Your Advances in Brain Imaging software is interfering. Use pathtool to move that course to the bottom.

[abdulkawi alareefi]

Thanks it worked.