MATLAB error referring to a function output variable that doesn't exist in the function

I'm getting this error

Output argument "varargout{2}" (and maybe others) not assigned during call to "DAGNetwork/predict".

when accessing the function predict in the code below

function YPredCell = yolov3Predict(net,XTrain,networkOutputs,anchorBoxMask)
% Predict the output of network and extract the confidence, x, y,
% width, height, and class.
YPredictions = cell(size(networkOutputs));
[YPredictions{:}] = predict(net, XTrain);
YPredCell = extractPredictions(YPredictions, anchorBoxMask);

% Apply activation to the predicted cell array.
YPredCell = applyActivations(YPredCell);
end

Here is the function predict. It has no varargout{2}.

function varargout = predict(this, varargin)
% predict   Make predictions on data with network

if this.NumInputLayers == 1
    % Network has one input
    X = varargin{1};
    nameValuePairs = varargin(2:end);
elseif iDataForMultipleInputNetworkPassedAsOneArgument(nargin-1, varargin) && ...
        iIsCombinedOrTransformedDatastore(varargin)
    % Network has multiple inputs, but the inputs are passed in as one
    % argument.
    X = varargin{1};
    nameValuePairs = varargin(2:end);
elseif iDataForMultipleInputNetworkPassedAsMultipleArguments(nargin-1, varargin, this.NumInputLayers)
    % Network has multiple inputs, and the inputs are passed in as multiple
    % arguments.
    X = varargin(1:this.NumInputLayers);
    nameValuePairs = varargin((this.NumInputLayers+1):end);
else
    error( message( 'nnet_cnn:DAGNetwork:InvalidPredictDataForMultipleInputNetwork', this.NumInputLayers) );
end

[options, executionEnvironment, acceleration, returnCategorical] = ...
    this.parseAndValidatePredictNameValuePairs( nameValuePairs{:} );

% Calculate predictions
try
    Y = this.calculatePredict( ...
        X, options, executionEnvironment, acceleration );
catch ex
    throw(ex);
end

% Convert outputs to categorical if necessary
if returnCategorical
    privateNetwork = this.PrivateNetwork;
    numOutputs = privateNetwork.NumOutputs;
    for i = 1:numOutputs
        outputLayer = privateNetwork.OutputLayers{i};
        if iIsClassificationOutputLayer(outputLayer)
            if ~this.NetworkInfo.LayerHasSequenceOutput( ...
                    privateNetwork.OutputLayerIndices(i) )
                Y{i} = iUndummify(Y{i}, outputLayer.Categories);
            else
                Y{i} = iUndummifySequence(Y{i}, outputLayer.Categories);
            end
        end
    end
end

varargout = Y;

end

Why am I getting the mentioned error when my function doesn't have the output argument?

[Stephen23]

"It has no varargout{2}."

Actually that depends entirely on the size of Y, which so far you have not told us.

Please show us the values of:

size(networkOutputs)
size(Y) % just before the function returns.

[Asim Shahzad]

@Stephen Cobeldick

size(Y)
ans =

     1     1
     
     
size(networkOutputs)
ans =

     2     1

How do I fix this?

Answer 1

[Walter Roberson]

YPredictions = cell(size(networkOutputs));

if network outputs passed in has size 2 or more then YPredictions will be a cell with two elements or more.

[YPredictions{:}] = predict(net, XTrain);

The left hand side is cell expansion and tells matlab that predict needs to return as many outputs as there are cell elements of YPredictions.

But your predict function only returns one output. And that is a problem if network outputs is size 2 or more.

[Asim Shahzad]

@Walter Roberson Ok, networkOutputs is 2x1 and predict returns a 1x1.

Any idea how to fix this? I'm following this tutorial.