Semi-Supervised Learning Algorithm Based on Lie Group
Hanxiang Xu, Fanzhang Li · 2009
Semi-supervised learning is an important research area in machine learning, which is mainly combined with a little labeled training data from reality, studies the data structure and distribution information from the large number of unlabeled data and makes full use of this information to improve the performance of classification algorithms, and researches the symmetry between the labeled and unlabeled samples. Lie Group is the combination of algebraic and geometrical structure by natural, it is the basic method to study the symmetry of the physical problems, so this paper introduces Lie Group to semi-supervised learning, analyzes the relationship between semi-supervised learning and Lie group, uses Lie group's nice algebraic and geometrical structure to denote and analyze data, gives the Semi-Supervised Learning algorithm based on Lie Group, and then in the experiment of predicting drug activity and comparing results with Self training, TSVM, and Co-training, shows the algorithm's feasibility and validity.