scikit-learn/scikit-learn

Missing Data Imputation with a Gaussian Mixture Model using EM

Open

#9,268 创建于 2017年7月3日

在 GitHub 查看
 (17 评论) (1 反应) (0 负责人)Python (66,084 star) (27,020 fork)batch import
help wantedmodule:mixture

描述

Description

I was wondering if there was interest in adding a new imputation strategy (or a new Imputer class) based on a Gaussian Mixture Model (GMM) using the EM or CEM algorithm. The implementation could be along the lines of:

  1. Ghahramani, Zoubin, and Michael I. Jordan. "Supervised learning from incomplete data via an EM approach." Advances in neural information processing systems. 1994.

  2. Ming Ouyang, William J. Welsh, Panos Georgopoulos; Gaussian mixture clustering and imputation of microarray data. Bioinformatics 2004; 20 (6): 917-923. doi: 10.1093/bioinformatics/bth007

  3. Machine Learning: A Probabilistic Approach, Kevin Murphy (2012) pp. 372-375. (although using a mixture rather than a single MVN in that example).

Further, Murphy's implementation of a Gaussian Mixture Model based imputation algorithm is also available here (which I have not tested, by the way).

EDIT: I am working on a completely revised approach, which is now based on reference 1 above. Will post a working version soon.

贡献者指南

技术栈
python
领域
machine learningdata
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
4
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
over 1 week
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
Gaussian Mixture ModelsEM algorithmPythonscikit learn internals
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
20
研究方向
This issue proposes implementing a new imputation strategy based on Gaussian Mixture Models (GMM) using the EM algorithm. The contributor should review the referenced papers by Ghahramani & Jordan, Ouyang et al., and Murphy's book, as well as the existing scikit learn imputation module (sklearn.impute) to understand integration. Additionally, the PMTK3 implementation of mixGaussImpute (https://github.com/probml/pmtk3/blob/master/toolbox/LatentVariableModels/mixGauss/mixGaussImpute.m) can serve as a reference. The issue has been stale since 2017, so the contributor should also check for any recent comments or PRs that may have addressed this.
Missing Data Imputation with a Gaussian Mixture Model using EM · scikit-learn/scikit-learn#9268 | Good First Issue