dlupdate

convGradientsFunction

TheconvGradientshelper function takes the learnable parameters of the convolution operation and a mini-batch of input datadlX，and returns the gradients with respect to the learnable parameters.

functiongradients = convGradients(dlX,params) dlY = dlconv(dlX,params.Weights,params.Bias); dlY = sum(dlY,'all'); gradients = dlgradient(dlY,params);end

Load Training Data

Load the digits training data.

[XTrain,TTrain] = digitTrain4DArrayData; classes = categories(TTrain); numClasses = numel(classes);

Define the Network

Define the network architecture and specify the average image value using theMeanoption in the image input layer.

layers = [ imageInputLayer([28 28 1],'Mean'，mean(XTrain,4)) convolution2dLayer(5,20) reluLayer convolution2dLayer(3,20,'Padding'，1) reluLayer convolution2dLayer(3,20,'Padding'，1) reluLayer fullyConnectedLayer(numClasses) softmaxLayer];

Create adlnetworkobject from the layer array.

net = dlnetwork(layers);

Define Model Loss Function

Create the helper functionmodelLoss，listed at the end of this example. The function takes adlnetworkobject and a mini-batch of input data with corresponding labels, and returns the loss and the gradients of the loss with respect to the learnable parameters.

Define Stochastic Gradient Descent Function

Create the helper functionsgdFunction，listed at the end of this example. The function takes the parameters and the gradients of the loss with respect to the parameters, and returns the updated parameters using the stochastic gradient descent algorithm, expressed as

where $\mathit{l}$是迭代数, $\alpha >0$is the learning rate, $\theta$is the parameter vector, and $\mathit{E}\left(\theta \right)$is the loss function.

Specify Training Options

Specify the options to use during training.

miniBatchSize = 128; numEpochs = 30; numObservations = numel(TTrain); numIterationsPerEpoch = floor(numObservations./miniBatchSize);

Specify the learning rate.

learnRate = 0.01;

Train Network

Calculate the total number of iterations for the training progress monitor.

numIterations = numEpochs * numIterationsPerEpoch;

Initialize theTrainingProgressMonitorobject. Because the timer starts when you create the monitor object, make sure that you create the object close to the training loop.

monitor = trainingProgressMonitor(Metrics="Loss"，Info="Epoch"，XLabel="Iteration");

Train the model using a custom training loop. For each epoch, shuffle the data and loop over mini-batches of data. Update the network parameters by callingdlupdatewith the functionsgdFunctiondefined at the end of this example. At the end of each epoch, display the training progress.

Train on a GPU, if one is available. Using a GPU requires Parallel Computing Toolbox™ and a supported GPU device. For information on supported devices, seeGPU Computing Requirements(Parallel Computing Toolbox).

iteration = 0; epoch = 0;whileepoch < numEpochs && ~monitor.Stop epoch = epoch + 1;% Shuffle data.idx = randperm(numel(TTrain)); XTrain = XTrain(:,:,:,idx); TTrain = TTrain(idx); i = 0;whilei < numIterationsPerEpoch && ~monitor.Stop i = i + 1; iteration = iteration + 1;% Read mini-batch of data and convert the labels to dummy% variables.idx = (i-1)*miniBatchSize+1:i*miniBatchSize; X = XTrain(:,:,:,idx); T = zeros(numClasses, miniBatchSize,"single");forc = 1:numClasses T(c,TTrain(idx)==classes(c)) = 1;end% Convert mini-batch of data to dlarray.X = dlarray(single(X),"SSCB");% If training on a GPU, then convert data to a gpuArray.ifcanUseGPU X = gpuArray(X);end% Evaluate the model loss and gradients using dlfeval and the% modelLoss function.[loss,gradients] = dlfeval(@modelLoss,net,X,T);% Update the network parameters using the SGD algorithm defined in% the sgdFunction helper function.updateFcn = @(net,gradients) sgdFunction(net,gradients,learnRate); net = dlupdate(updateFcn,net,gradients);% Update the training progress monitor.recordMetrics(monitor,iteration,Loss=loss); updateInfo(monitor,Epoch=epoch +" of "+ numEpochs); monitor.Progress = 100 * iteration/numIterations;endend

Test Network

Test the classification accuracy of the model by comparing the predictions on a test set with the true labels.

[XTest,TTest] = digitTest4DArrayData;

Convert the data to adlarraywith the dimension format"SSCB"(spatial, spatial, channel, batch). For GPU prediction, also convert the data to agpuArray.

XTest = dlarray(XTest,"SSCB");ifcanUseGPU XTest = gpuArray(XTest);end

To classify images using adlnetworkobject, use thepredictfunction and find the classes with the highest scores.

YTest = predict(net,XTest); [~,idx] = max(extractdata(YTest),[],1); YTest = classes(idx);

Evaluate the classification accuracy.

accuracy = mean(YTest==TTest)

accuracy = 0.9040

Model Loss Function

The helper functionmodelLosstakes adlnetworkobjectnetand a mini-batch of input dataXwith corresponding labelsT，and returns the loss and the gradients of the loss with respect to the learnable parameters innet. To compute the gradients automatically, use thedlgradientfunction.

function[loss,gradients] = modelLoss(net,X,T) Y = forward(net,X); loss = crossentropy(Y,T); gradients = dlgradient(loss,net.Learnables);end

Stochastic Gradient Descent Function

The helper functionsgdFunctiontakes the learnable parametersparameters，the gradients of the loss with with respect to the learnable parameters, and the learning ratelearnRate，and returns the updated parameters using the stochastic gradient descent algorithm, expressed as

where $\mathit{l}$是迭代数, $\alpha >0$is the learning rate, $\theta$is the parameter vector, and $\mathit{E}\left(\theta \right)$is the loss function.

functionparameters = sgdFunction(parameters,gradients,learnRate) parameters = parameters - learnRate .* gradients;end