Estimation of functional dependency parameters at zero median value of noises in measurements

In the paper, we study a problem of estimating functional dependency parameters by empirical data with random noises. A specific feature of the formulation is a replacement of a traditional assumption of zero mathematical expectation of a noise by the condition of equality to zero of a median value of the noise. We represent an algorithm for constructing estimates of parameters minimizing a mathematical expectation of losses including the variant of spline approximation. We interpret such estimates. The results of the experiments are given.