Integration of machine learning and human learning for training optimization in robust linear regression

Abstract

In this paper machine learning and human learning are applied jointly to optimize the training of linear regression. Human learning is exploited to label extra training data so as to resolve problems such as insufficient training and over-fitting. Considering the inevitable human errors in labeling, two machine learning algorithms are developed which optimize the selection of the extra training data and detect human errors during linear regression. The first algorithm assumes sparse human errors and implements a sparse optimization within a sequential active learning procedure. The second algorithm deals with non-sparse human errors. By exploiting the IRT (item response theory) to model the distribution of human errors, it reconstructs the training data set so that the human labeling errors become sparse. Simulations are conducted to show that the two algorithms are effective in resolving the insufficient training and human labeling error problems.