Experiment Manager Setup for LSTM

I want to optimize a LSTM network using the Experiment Manager.

The training data XTrain consists of a cell array. Each cell contains a matrix of type double. The classification is stored in YTrain as a categorical array.

I tried to store the TrainingData in an cell array containg XTrain and YTrain.

function [TrainingData, layers, options] = Experiment1_setup3(params)

load('LSTM_dataset.mat', 'XTrain', 'YTrain', 'XTest', 'YTest')

TrainingData = cell(length(YTrain),1);

for i = length(YTrain)

TrainingData{i} = {XTrain{i} YTrain(i)};

end

inputSize = length(Xtrain{1}(:,1));

numHiddenUnits = 100;

numClasses = length(categories(YTrain));

options = trainingOptions('adam', ...

'ExecutionEnvironment',"auto", ...

'GradientThreshold',1, ...

'MaxEpochs', params.maxEpochs, ...

'MiniBatchSize', params.miniBatchSize, ...

'InitialLearnRate', 0.001, ...

'SequenceLength','longest', ...

'Shuffle','never', ...

'Verbose',0);

I got the following error message:

Error(s) occurred while validating the setup function:

Caused by:

Unable to determine if experiment is for classification or regression because setup function returned invalid outputs.

Not enough input arguments.

Error using nnet.internal.cnn.trainNetwork.DLTInputParser>iParseInputArguments (line 94) Not enough input arguments.

I want to ask how to prepare the data for the Experiment Manager since all examples I found use images.

Thank you

[Marco Lutz]

I found the error. I simply have the change the first line to:

function [XTrain, YTrain, layers, options] = Experiment1_setup3(params)

For optimization of the XTest prediction i had to create a new function following this example https://de.mathworks.com/help/deeplearning/ug/experiment-using-bayesian-optimization.html

I found it untypically difficult for Matlab to find the correct information on this matter.

Answer 1

[Jorge Calvo]

Hi Marco,

This example documents the various syntaxes you can use for the experiment setup function: https://www.mathworks.com/help/deeplearning/ref/experimentmanager-app.html#mw_92d8f99c-f283-4bc5-a5d7-7d779c4831f7

It sounds like you discovered the fourth syntax, hopefully without too much pain!

By the way, you may also be interested in this example, since it involves training an LSTM: https://www.mathworks.com/help/deeplearning/ug/exp-mgr-sequence-regression-example.html

Cheers,

Jorge