Hi I am trying to run GP regression on mixed type of data. Specifically, other than normal data that can tell us there the output variable y_i at input variable x_i, I have two other types observations that can only inform me that y_i at x_i should larger than a value or lower than a value. In other words, they are ‘limiting point’ with uniform distribution of [-inf, y_i] or [y_i,inf].
In this case, it is possible to express three two types of likelihood function (i.e., Gaussian and Uniform) for GP regression?