deep learning for regression in matlab

main.m
% Load data from a CSV file
data = readtable('data.csv');
inputs = data(:, 1:end-1);
targets = data(:, end);
109 chars
5 lines

main.m
% Split data into training, validation, and testing sets
[trainInd,valInd,testInd] = dividerand(size(inputs,1), 0.6, 0.2, 0.2);
xTrain = inputs(trainInd,:);
yTrain = targets(trainInd,:);
xVal = inputs(valInd,:);
yVal = targets(valInd,:);
xTest = inputs(testInd,:);
yTest = targets(testInd,:);
293 chars
9 lines

main.m
% Define the architecture of the neural network
numFeatures = size(inputs, 2);
numResponses = size(targets, 2);
numHiddenUnits = 100;
layers = [
    sequenceInputLayer(numFeatures)
    fullyConnectedLayer(numHiddenUnits)
    reluLayer()
    fullyConnectedLayer(numHiddenUnits)
    reluLayer()
    fullyConnectedLayer(numResponses)
    regressionLayer()];
355 chars
13 lines

main.m
% Specify the training options
options = trainingOptions('adam', ...
    'MaxEpochs', 200, ...
    'MiniBatchSize', 128, ...
    'LearnRateSchedule', 'piecewise', ...
    'LearnRateDropFactor', 0.1, ...
    'LearnRateDropPeriod', 50, ...
    'ValidationData', {xVal, yVal}, ...
    'ValidationFrequency', 10, ...
    'Plots', 'training-progress');
348 chars
11 lines

main.m
% Train the neural network
net = trainNetwork(xTrain, yTrain, layers, options);
80 chars
3 lines

main.m
% Make predictions on the testing set
ypred = predict(net, xTest);

% Evaluate the performance of the model
MSE = mean((ypred - yTest).^2);               % Mean Squared Error
RMSE = sqrt(mean((ypred - yTest).^2));        % Root Mean Squared Error
R2 = 1 - sum((ypred - yTest).^2)/sum((yTest - mean(yTest)).^2);  % R-squared
324 chars
8 lines