Regression-type problems under zero median noise

A study was conducted to demonstrate regression-type problems under zero median noise. An algorithm was constructed for estimating parameters of a given dependence under the assumption that noise had median zero, which was less vulnerable. The algorithm estimated parameters of a given dependence, including a spline approximation that minimized empirical risk. The algorithm for constructing the function P0(THETA) was based on constructing a system of pairwise disjoint open sets whose union had closure coinciding with D so that the probability that a vector ΘT belonging to each of these sets be estimated from the given data.