APP DESIGNER: How to convert value input by user into array and test it using trained neural network?

Question

Nurliana Farhana Salehuddin 2022-2-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1658090-app-designer-how-to-convert-value-input-by-user-into-array-and-test-it-using-trained-neural-network

评论： Cris LaPierre 2022-2-25

采纳的回答： Cris LaPierre

在 MATLAB Online 中打开

Dear MATLAB user,

I have been stucked with this problem for several days and I could not figure out what are my mistakes?

The aim of my app designer is to receive input from user > predict using trained neural network > show output at GUI

Problem that I face now is to convert the input variables (3 variables) by user into an array. This is because I used an array in the network. The code that I used:

 T=table(app.Variable1.Value,app.Variable2.Value,app.Variable3.Value,);
 T2=table2array(T);

but somehow I dont think this is the right way to do this.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Cris LaPierre 2022-2-25

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1658090-app-designer-how-to-convert-value-input-by-user-into-array-and-test-it-using-trained-neural-network#answer_904090

编辑：Cris LaPierre 2022-2-25

在 MATLAB Online 中打开

Why create a table just to turn it back into a matrix? Why not just concatenate the 3 variables? Since we know nothing about what the input looks like, here is a general suggestion

T2=[app.Variable1.Value,app.Variable2.Value,app.Variable3.Value];

You may need to transpose your variables depending on how the user inputs them to get your array to be organized the way your network expects.

2 个评论
显示无隐藏无

Nurliana Farhana Salehuddin 2022-2-25

编辑：Nurliana Farhana Salehuddin 2022-2-25

在 MATLAB Online 中打开

Dear @Cris LaPierre,

Thank you. I also dunno why I make it complicated.

Would like to ask for your opinion, I use mapstd to scale my input for the trained network.

x = TrainInput; 
t = TrainOutput;
[x1 ps]=mapstd(x);
[t1 ts]=mapstd(t);

Thus, in app designer I also need to convert the input before using the network right? But how do I convert back the mapstd values after prediction into the original scale ?

Cris LaPierre 2022-2-25

You need to preprocess your actual data the exact same way you preprocessed your training data for your model to give meaningful results.

The way to get back to original values is to do the opposite of what mapstd is doing. You can see the algorithm used here. Just a note that I would not expect to get exactly the same values back, but they should be fairly close. You will need to use the mean and std values you passed into mapstd.

请先登录，再进行评论。