Create a package for performance function in Matlab "perform.Fcn" from scratch without building above existed one

Hi everyone

I am trying to create a shallow NN that estimates the parameters of an equation given the input and output. The perfromance function is using the parameters to estimate y_pred and then compare it with y_act and back propogated the error. I need to create a package doing the same thing but without building on existing one like mse because I didn't know how to tune it according to this question:

https://www.mathworks.com/matlabcentral/answers/461848-custom-performance-function-for-shallow-neural-networks-using-mse-package

So I want to built one from scratch. To make the idea clearer I want to estimate the parameters using NN in away that is similar to RLS, that the parameters are adjusted using the error between the estimated output (Using the parameters) and the actual output

[Steven Lord]

Do you have a specific question about implementing a new performance function? I see that you reopened this question. I suspect it was closed initially because it wasn't clear what you were looking for from the Answers community.

[Baker Shehadeh]

For closing the question, I wrote an email to clarify that it didn't break any rule so I reopened it. For implementing a new performance function, I tried a lot of different mehtods to create a custom function and assign it to net.perform.Fcn but it didn't work because Matalb only accepts the existed one. I went throught your Q/A to fnd what I need and I found this

https://www.mathworks.com/matlabcentral/answers/461848-custom-performance-function-for-shallow-neural-networks-using-mse-package but it didn't help because I want to pass for the function that I want to customized a lot of data not only y_act and y_pred so I want to know how to create one from scratch not building upon existing one because it will not help. So how to build a package for my customized function and implemented and assign it to net.perfrom.Fcn ?