A good tutorial of Solve Partial Differential Equations Using Deep Learning (physics informed neural networks)

Hello, instead of Python, I want to learn physics informed neural networks in MATLAB. I am using R2020a

https://www.mathworks.com/help/deeplearning/ug/solve-partial-differential-equations-using-deep-learning.html

Like there is no function called initializeHe, initializezeros, etc

I found this tutorial, but I just cannot follow it. What I suppoesed is to have a complete code, then I can make breakpoints, etc to learn it. However this one too many jumps, just don't know where I should look.

Can anyone help me?

Answer 1

[David Willingham]

What version of MATLAB are you using? There is a function called arraydatastore.

Did you try opening the full script in MATLAB and running from there?

>> openExample('nnet/TrainPhysicsInformedNeuralNetworkWithDLTExample')

[Miraboreasu]

Thank you for the reply, I am using R2020a, and for the openExample, it said not found in my computer.

[per isakson]

It was introduced in R2020b. See arrayDatastore - Datastore for in-memory data

[Miraboreasu]

Then how about the function initializeHe, initializeZeros etc

[Joseph Eom]

https://www.mathworks.com/help/deeplearning/ug/solve-partial-differential-equations-using-deep-learning.html

This example supporting R2021a. I think you should upgrade recent version of MATLAB. initializeHe,initializeZeros is internal function to run this example.

Answer 2

[SEYED JAMALEDDIN MOSTAFAVI YAZDI]

I had same issue. I resolved it by installing the deep learning toolbox of MATLAB.

https://www.mathworks.com/products/deep-learning.html

Answer 3

[Akam Aboubakri]

I had the same problem. I solved it with adding the following functions:

function parameter = initializeHe(sz,numIn,className)
arguments
    sz
    numIn
    className = 'single'
end
parameter = sqrt(2/numIn) * randn(sz,className);
parameter = dlarray(parameter);
end

function parameter = initializeZeros(sz,className)
arguments
    sz
    className = 'single'
end
parameter = zeros(sz,className);
parameter = dlarray(parameter);
end