Hi,
I would like to know how we can construct a neural network that for a given set of inputs gives a vector y depending on vector x. In this example both vectors x and y are put in the output part of the network. A trivial example would be to have a gaussian function as output for a given set of (mu,sigma) in input:
(mu_1,sigma_1) -> gaussian y_1 as function of x_1
(mu_2,sigma_2) -> gaussian y_2 as function of x_2
...
(mu_n,sigma_n) -> gaussian y_n as function of x_n
In this example, the range of x-values would differ for each data set. To create the x vector, one can set x_min equal to the minimum of x_1(1),...,x_n(1), and x_max equal to the maximum of x_1(end),...,x_n(end). However, in my real test case, I don't work with an analytical function, and creating a general x vector would require adding zeros to the y-values for values less than the x_i(1) and greater than x_i(end). Asking the network to correctly predict these zero values is neither wise nor time-efficient.
Thanks in advance for your help
