LSTM Hysteresis curve modeling

Hello,

I need to model a hysteresis curve by LSTM neural network can anyone help me with that?

I've read matlab documentations but unf I could'nt find my case. In face I don't know how to define properly network architucture (layer's).

Thanks for your help

[yanqi liu]

yes，sir，may be upload your data mat file to analysis

[behrad rashedi]

it is two columns of simple numbers as input and output.

Answer 1

[Milan Bansal]

Hi behrad,

It is my understanding that you want to build a Long Short-Term Memory (LSTM) Network to model a hysteresis curve using your data.

Below are high level steps to build an LSTM Model to model the hysteresis curve:

1. Preprocess Data: Normalize the data if necessary and structure it into sequences that the LSTM can process.
2. Create a sequence-to-sequence LSTM network architecture using Deep Learning Toolbox. The design includes choosing the number of LSTM layers, the number of hidden units in each layer, and other hyperparameters.
3. Train the LSTM network using the training data.

Please refer to the pseudo code in the code snippet below: -

% Normalize data (if necessary) 
inputDataNorm = normalize(inputData); 
outputDataNorm = normalize(outputData); 

% Prepare sequences - t 
XTrain = ... % Your input sequences 
YTrain = ... % Your output sequences 
    
% Define LSTM network architecture 
layers = [ 
    sequenceInputLayer(size(XTrain,1)) 
    lstmLayer(50,'OutputMode','sequence') 
    fullyConnectedLayer(1) 
    regressionLayer 
    ]; 

% Specify training options 
options = trainingOptions('adam', ... 
    'MaxEpochs',100, ... 
    'GradientThreshold',1, ... 
    'InitialLearnRate',0.005, ... 
    'LearnRateDropPeriod',125, ... 
    ); 

% Train LSTM network 
net = trainNetwork(XTrain, YTrain, layers, options); 

Please refer to the documentation below to learn more about LSTM Network.

https://www.mathworks.com/help/deeplearning/ug/long-short-term-memory-networks.html

Please refer to the documentation below to learn more about "lstmLayer".

https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.lstmlayer.html

Hope it helps!