基于熵理论的单类学习机

摘要

One-class classification is widely concerned by the researchers .A one-class learning machine based on entropy ( OLME) is proposed in this paper after analysing the current mainstream methods .OLME takes the data distributions into full consideration and uses entropy to stand for the uncertainties of the classification .The effectiveness of the classification is guaranteed by minimising the entropy .The experiments on standard datasets demonstrate that compared with current mainstream one -class classification methods , OLME has higher classification efficiency .%单类问题受到业界广泛关注。在分析当前主流单分类方法基础上，提出基于熵理论的单类学习机OLME （ One-class Learning Machine based on Entropy ）。该方法充分考虑了数据的分布性状，利用熵来表征分类的不确定性，通过最小化熵保证分类的有效性。标准数据集上的实验表明，与当前主流的单分类方法相比，OLME具有较高的分类效率。