Fitting fields/functions with NN

[lucifirm]

I know I can use a MLP to fit a function, but I have a couple of questions:

1) What is the simplest NN that I could use for fitting an unknown vector field, let's say an electric and/or magnetic field in a room (when I don't know the source(s), of course)?

2) What is the best NN that I could use for fitting the same vector field? Can I use one NN for fitting and other for optimizing the first NN?



P.S.: I'm new to the subject and I'm taking the LM course...

[htlin]

Originally Posted by lucifirm

I know I can use a MLP to fit a function, but I have a couple of questions:

1) What is the simplest NN that I could use for fitting an unknown vector field, let's say an electric and/or magnetic field in a room (when I don't know the source(s), of course)?

2) What is the best NN that I could use for fitting the same vector field? Can I use one NN for fitting and other for optimizing the first NN?



P.S.: I'm new to the subject and I'm taking the LM course...

In general, there is no "best" until you can define what your goal is. In general, people use the same model (say, NN) to learn from the data (optimizing/fitting) and for making predictions, and may need to choose among several possibilities (say NN structure) using learning tools (such as validation illustrated in Chapter 4 of the book). Hope this helps.