Regression Equation from artificial neural network

Hi, I have 3 Input Data (temperature, moisture content, and frequency) set of 168 and

my target data (dielectric constant) set of 168.

I used an artificial neural network toolbox to get a regression model to generate simulated data.

I trained(70%), validated (15%), and tested(15%) it. I got reasonable result.

Once i got the result is it possible to extract the regression equation from result.

If it is possible, which section should I click to get the regression model??

If I am not able to get the regression model directly, how can I get other information like bias, weight, and structure of neural network model to generate the regression model in directly or manually?

Answer 1

[Greg Heath]

The default regression equation between inputs and outputs is a curve in 3-dimensional input space.

 y = B2 +LW*tansig(B1+IW*x),

where the weights are obtained given the target, t.

The plots you have are the 1-D regressions of output vs target.

 y = W*t + b;

However, IW, B1, B2 and LW cannot be obtained by using W and b.

IW = net.IW; b = net.b; LW = net.LW;

Hope this helps.

Thank you for formally accepting my answer

Greg

[daeung Yu]

Thanks for quick answering.

I'm wondering y = B2 +LW*tansig(B1+IW*x)is this the default equation for artificial neural network.

That mean whenever I use neural network, I can use that default equation? or only for when I use the tansig function??

And how can I get a IW, b, and LW values??

Thanks

[Greg Heath]

>I'm wondering y = B2 +LW*tansig(B1+IW*x)is this the default equation for artificial neural network.

 1. For fitnet, patternnet, feedforwardnet and obsolete predecessors
 2. Provided the default normalization is taken into account.

>That mean whenever I use neural network, I can use that default equation? or only for when I use the tansig function??

 See above

> And how can I get a IW, b, and LW values??

 See above.

[Greg Heath]

Check out the 14 threads obtained when searching on

greg IW LW b b1 b2

[daeung Yu]

I got it

Thanks for your sincere help.

[praveen kumar ramagiri]

Hi Greg,

Your answer is helpful but i'm unable to get it clearly.

Here, what is x? how will i get its value?

Thank You in advance.

Answer 2

[Kachukwu Okoh]

Hello Daeung Yu, Hello Greg, good day from over here. Sorry to bring your attention back to this. I am working on a project of this sort of example you gave where I now have to use the model gotten to predict the outcome of new input data of same aspect.

Please is there any way (book or article) you could share with me to help me understand fully how to go about it. I.e. from making necessary changes (activation functions, learning rules, etc) for improving the model with neural network toolbox to extracting the model for prediction of new input data?

Thanks in anticipation. 🙏🏼