Issue with LSTM training

% Using the LSTM NN function.

clc; clear all; close all;

% Training the function.

inputTrain = xlsread('Vehicle_Data_MY19CadillacCT6.xlsx',1,'C38:L100');

ouputTrain = xlsread('Vehicle_Data_MY19CadillacCT6.xlsx',1,'M38:O100');

% Testing the results

inputTest = xlsread('Vehicle_Data_MY19CadillacCT6.xlsx',1,'C101:L110');

ouputTest = xlsread('Vehicle_Data_MY19CadillacCT6.xlsx',1,'M101:O110');

inputSize = 10;

outputSize = 3;

numHiddenUnits = 50;

layers = [ sequenceInputLayer(inputSize) lstmLayer(numHiddenUnits)

fullyConnectedLayer(outputSize) regressionLayer];

options = trainingOptions('adam', ...

'MaxEpochs',1000,...

'GradientThreshold',0.01, ...

'InitialLearnRate',0.0001);

net = trainNetwork(inputTrain,ouputTrain,layers,options);

outputPrediction = predict(net,inputTest);

Error using trainNetwork (line 170)

Error setting property 'ExternalLayers' of class 'nnet.internal.cnn.analyzer.NetworkAnalyzer':

Size of value must match specified dimensions M×1.

Error in PhD_Thesis_1 (line 24)

net = trainNetwork(inputTrain,ouputTrain,layers,options);

Caused by:

Error using nnet.internal.cnn.layer.util.inferParameters (line 7)

Error setting property 'ExternalLayers' of class 'nnet.internal.cnn.analyzer.NetworkAnalyzer':

Size of value must match specified dimensions M×1.

Answer 1

[Pranav Verma]

Hi Srikant,

From the provided code, it seems that you are creating an LSTM network for regression and while defining the layers, inputSize and outputSize does not match with the inputTrain and ouputTrain sizes. While defining the sequenceInputLayer size and fullyConnectedLayer size, use the sizes directly from the inputTrain and ouputTrain using the size function. Please refer to below documentation for size function:

https://www.mathworks.com/help/matlab/ref/size.html

Also, please refer to the following example for creating a regression LSTM network.

https://www.mathworks.com/help/deeplearning/ug/sequence-to-sequence-regression-using-deep-learning.html

Thanks,

Pranav Verma