Radial Basis Function with "newrb" dimensional problem

I'm trying to use the function "newrb" for generating an output considering a set of initial inputs and the RBF.

My problem is that as input x I have 5 set of 8 elements (8 coordinates in a 8D space), while as output y I would like to obtain just 1 value (the corresponding 9th coordinates). However, the function said that y must have the same dimension of the value used for x.

To better explain myself, if we consider just 1 set, which has 8 coordinates, I want to use the neural network for obtaining just 1 final value, but I am not allowed to do this since "newrb" required an y with the same dimension of x.

How can I solve this issue??

Answer 1

[nick]

Hi @Francesco Porretta,

The "newrb" function requires the target matrix T to have the same number of columns as the input matrix P. This means that if your input matrix P has dimensions [R x Q], your target matrix T should have dimensions [S x Q], where Q is the number of samples.

The following MATLAB code shows RBF network to predict a single value (the 9th coordinate) for 5 sets using 8 elements (8 coordinates in an 8D space) of each set:

% Sample input data 
P = rand(8, 5); % 5 sets of 8 elements (8 coordinates in an 8D space)

% Sample output data 
T = rand(1, 5); % 9th coordinate for each set

net = newrb(P, T);
view(net);

You may refer to the following documentation to learn more about "newrb" function :

https://www.mathworks.com/help/releases/R2021a/deeplearning/ref/newrb.html