This is a so-called Forward Stage-wise Additive Modelling which appears in the course "FOUNDATIONS OF MACHINE LEARNING" lecture 23. https://bloomberg.github.io/foml/#lectures
This repository shows an example, which is coded by myself using C++, about using FSAM and decision stumps to do the regression on some sample datas.
In header file "randompoint.h", there is an function called quadratic which can be used to generate some points, and each point follows f(x) = x^2 + N(mean, sigma).
To run the program, you only need to call the Loop function in "boosting.h"
Note in order to do the splitting properly, you need to sort the data first. I skipped this procejure because the points are already in an ascent order.