A METHOD FOR ADJUSTING THE FIRING VECTOR WHEN BUILDING PIECEWISE LINEAR REGRESSION MODELS

The paper provides a brief overview of publications related to the construction and analysis of nonlinear regression models. Specifically, it considers the following topics: a piecewise linear regression construction method based on a new formulation of the integer linear programming problem, a methodology for constructing interpretable models of the structure-activity chemical bond by combining elements of network analysis and piecewise linear regression, an extension of the piecewise linear model construction method that solves the problem of choosing the optimal number of partitions and retraining, and an algorithm for constructing automatic piecewise linear regression that combines the predictive capabilities of reinforcement learning with the interpretability of multidimensional adaptive regression splines. It also investigates popular piecewise linear regression methods, such as the Leontief and risk functions. An important attribute of these models is the so-called response vector...