After load ANN data in App Designer, how to predict from any input from user, then show the output? (callback is applied)

My ANN coding is first being run in the matlab .m file before loaded into App Designer.

The neural network data has two input columns (A and B) and one output column (C). (Refer to first picture)

After run in matlab, the output shown in command window is in two columns (refer to the second picture).

An app designer is made (refer to third picture) to present this neural network better and the data is loaded succesfully. However, to make the 'predict' button works, the coding for prediction keeps showing error (refer to the last picture). Tried a lot of things and the error is always for this prediction line only. This error pops up after two inputs are inserted (as number) and the predict button is pressed. Callback is made for predict button only.

Any suggestion to solve this problem?

The coding for 'predict' button:

% Button pushed function: PredictButton

function PredictButtonPushed(app, event)

% Get user input from edit fields

InputRating = str2double(app.InputRatingEditField.Value);

InputCurrent = str2double(app.InputCurrentEditField.Value);

%%inputTable = table(InputRating, InputCurrent);

% Create a matrix with the input data

inputData = [InputRating, InputCurrent];

% Perform prediction using your ANN model

%%prediction = predict(app.net,InputRating,InputCurrent);

prediction = predict(app.net, inputData);

% Display the prediction result in the output text

app.OutputText.Value = sprintf('Prediction: %.2f', prediction);

% Display the prediction result in the output text

app.OutputRatingEditField.Value = string(prediction);

end

end

[Venu]

Hi @Aida Zulaikha,

Can you provide some information on your ANN model for clear understanding of the issue.

Answer 1

[Cris LaPierre]

To make predictions, I suggest using your model like a function, meaning you pass the input to your model directly.

prediction = app.net(inputData)

[Aida Zulaikha]

Hello, thank you for your answer.

I've tried your suggestion and got another error which is:

Here are the coding for my neural network model:

% Input layer
x = [BreakerRating, Current];

% Output layer
t = OutputBreakerRating;

x = x';
t = t';

% create feedforward network
 net = feedforwardnet([10,2],'trainbr');
 
% Specify activation function
net.layers{1}.transferFcn='tansig';     % First Hidden Layer
net.layers{2}.transferFcn='logsig';     % Output layer

net = configure(net,x,t);               % To run input & output

view(net);                              % To view feedforward neural network

% To divide data randomly for train, validation and test
net.divideFcn='dividerand';
net.divideParam.trainRatio = 0.75;
net.divideParam.valRatio = 0.10;
net.divideParam.testRatio = 0.15;

% to setup training parameter
net.trainParam.epochs = 100;        % No of Iteration
net.trainParam.goal = 1e-4;         % Error set
net.trainParam.lr = 0.01;           % Learning rate

net = train(net,x,t);               % Command to train neural network

Predicted_output = net(x);          % Predict output

Predicted_output = Predicted_output';   % change horizontal data to vertical

t = t';                             % Change horizontal data to vertical

disp([t,Predicted_output]);         % Compare target and output

plotregression(t,Predicted_output)  % To observe regression

save('breakersize2.mat','net');

Attached are the .csv data for running the neural network.

The first line to the callback buttons coding for the app design is:

classdef breakersize < matlab.apps.AppBase

    % Properties that correspond to app components
    properties (Access = public)
        UIFigure                    matlab.ui.Figure
        InputRatingEditField        matlab.ui.control.EditField
        InputRatingEditFieldLabel   matlab.ui.control.Label
        OutputLabel                 matlab.ui.control.Label
        InputCurrentEditField       matlab.ui.control.NumericEditField
        InputCurrentEditFieldLabel  matlab.ui.control.Label
        StatusLabel                 matlab.ui.control.Label
        PredictButton               matlab.ui.control.Button
        LoadANNButton               matlab.ui.control.Button
    end

    
    properties (Access = private)
        filepath % file path
        net % Trained Neural Network
    end
    
% Callbacks that handle component events
    methods (Access = private)

        % Button pushed function: LoadANNButton
        function LoadANNButtonPushed(app, event)
        % Load the trained model
        app.net = load('breakersize2.mat');

        % Display a message indicating that the model is loaded
        disp('Trained model loaded successfully.');

        % Update a text label in the app interface
        app.StatusLabel.Text = 'Trained model loaded successfully.';
        end

        % Button pushed function: PredictButton
        function PredictButtonPushed(app, event)
        % Get user input from edit fields
        InputRating = str2double(app.InputRatingEditField.Value);
        InputCurrent = str2double(app.InputCurrentEditField.Value);

        % Create a matrix with the input data
        inputData = [InputRating, InputCurrent];

        % Perform prediction using your ANN model
       %prediction = predict(app.net, InputData);
       prediction = app.net(inputData);

        % Display the prediction result in the output text
        app.OutputText.Value = sprintf('Prediction: %.2f', prediction);
        % Display the prediction result in the output text
        app.OutputRatingEditField.Value = string(prediction);
        end
    end

I don't know how to solve this problem since i'm a beginner for app designer.

[Cris LaPierre]

For app designer, my suggestion is to use an Edit Field (Numeric) when you are collecting or displaying numbers. Then you don't have to worry about using str2double or string.

Other than that, your app designer code looks correct. You've correctly created app properties, and you've added most of the components your code uses. I only noticed that, based on your code, there is no OutputText component. You might have changed the name of your label, so it might still work for you.

For labels, the display property is Text, not Value

Also, I didn't see an OutputRatingEditField in your public app properties. Make sure you add this component or your code will return an error.

When you load a mat file using the syntax you are using in your app - S = load(___) - it loads the variable(s) to a structure. So app.net is a structure, not a network. You must go one more level to get to your network: app.net.net

Finally, notice in your first code, your input to net is x, a 2x480 array. So row 1 is the first input value, and row 2 is the 2nd input. Each column is a separate prediction. Make sure your inputs are a column vector, not a row vector (use a semicolon to separate then instead of a comma).

I only had to make changes to your Predict button callback to build a working app.

function PredictButtonPushed(app, event)
    % Get user input from edit fields
    InputRating = app.InputRatingEditField.Value;  % removed str2double
    InputCurrent = app.InputCurrentEditField.Value;

    % Create a matrix with the input data
    inputData = [InputRating; InputCurrent]; % turned into a column vector

    % Perform prediction using your ANN model
    %prediction = predict(app.net, InputData);
    prediction = app.net.net(inputData);   % access desired field in structure

    % Display the prediction result in the output text
    app.OutputLabel.Text = sprintf('Prediction: %.2f', prediction); % assign to Text property

    % Display the prediction result in the output text
    app.OutputRatingEditField.Value = prediction; % removed string
end

[Aida Zulaikha]

Hello, I followed your recommendations and that works smoothly. Thank you very much for the help!

I appreciate your time and effort a lot! Thank you!