这个劳动节,本人发布了用于估计copula熵的Python包 copent 。copula熵是一种为多变量统计相关定义的数学概念[1],本包实现了copula熵的非参数估计方法,可以不加假设地应用于任何情况。copula熵已被用于关联分析[2](代替皮尔逊相关系数),图结构学习[3],变量选择[4](代替lasso,AIC,BIC,distance correlation,HSIC等),因果发现[5](通过估计传递熵Transfer Entropy)等等,具体见相应参考文献。
copent包可以pip直接安装:
pip install copent
其源码在github上共享:
https://github.com/majianthu/pycopent/
同名R包copent已放在CRAN上,
CRAN: https://cran.r-project.org/package=copent
GITHUB: https://github.com/majianthu/copent/
相关参考文献均已上传至arXiv:
arxiv.org/a/ma_j_3
希望对大家有所帮助。
------
References
1. Ma Jian, Sun Zengqi. Mutual information is copula entropy. Tsinghua Science & Technology, 2011, 16(1): 51-54. See also arXiv preprint, arXiv:0808.0845, 2008.
2. Ma Jian. Discovering Association with Copula Entropy. arXiv preprint arXiv:1907.12268, 2019.
3. Ma Jian, Sun Zengqi. Dependence Structure Estimation via Copula. arXiv preprint arXiv:0804.4451v2, 2019.
4. Ma Jian. Variable Selection with Copula Entropy. arXiv preprint arXiv:1910.12389, 2019.
5. Ma Jian. Estimating Transfer Entropy via Copula Entropy. arXiv preprint arXiv:1910.04375, 2019.