PRML Chapter 9. Mixture Models and EM
2012-03-07 16:48
218 查看
今天从网上搜的 EM 算法的原始论文叫做 Maximum likelihood from incomplete data via the EM algorithm,下次仔细学习的时候可以看看,最近比较功利,就不弄得那么明白了。(2012@3@21)
接下来第二小结是 Mixtures of Gaussians,看来不得不回2.3去看了。
第一部分链接:Mixtures of Gaussians
第二部分链接:9.2.1 Maximum likelihood
9.2.2 的推导看不明白,直接上结论
1. 对于GMM的EM算法,由于EM计算量大而且收敛慢,可以先用K-means找到几个中心,然后均值、方差和混合参数都可以根据K-means的结果设定初始值。
2. 需要防止模型在奇异点崩塌
3. EM不能保证找到全局最优值
下面不打字了,直接把书上截图了:
9.1 K-means Clustering
主要介绍了 K-means 和 EM 算法之间的关系,第一次听说原来 K-means 就是 EM 算法,不知道的东东还真是多。接下来第二小结是 Mixtures of Gaussians,看来不得不回2.3去看了。
9.2 Mixtures of Gaussians
这章内容太多,所以单独写成分日志。第一部分链接:Mixtures of Gaussians
第二部分链接:9.2.1 Maximum likelihood
9.2.2 的推导看不明白,直接上结论
1. 对于GMM的EM算法,由于EM计算量大而且收敛慢,可以先用K-means找到几个中心,然后均值、方差和混合参数都可以根据K-means的结果设定初始值。
2. 需要防止模型在奇异点崩塌
3. EM不能保证找到全局最优值
下面不打字了,直接把书上截图了:
相关文章推荐
- PRML Chapter 9. Mixture Models and EM
- MLaPP Chapter 11 Mixture models and the EM algorithm
- PRML读书会第九章 Mixture Models and EM(Kmeans,混合高斯模型,Expectation Maximization)
- Chapter 9: Mixture Models and EM
- PRML读书会第九章 Mixture Models and EM(Kmeans,混合高斯模型,Expectation Maximization)
- 第九章:Gaussian Mixture and EM exercise 1-14
- 第九章:Gaussian Mixture and EM exercise 15-27
- ML 12 & 13 mixture of gaussions and EM
- Infinite Mixture Models with Nonparametric Bayes and the Dirichlet Process
- dango models and database ---- relation ship
- Database and models
- [Machine Learning] Probabilistic Graphical Models:一、Introduction and Overview(2、Factors)
- mental ray feat. “Physically Based Shading Models in Film and Game Production: Practical Implementation at tri-Ace"
- 动态目标检测(一):Adaptive Background Mixture Models for Real-Time Tracking
- [导入]generic file and image models for uploaded files
- Discrimination Models , Generative Models and Naive Bayes
- Classification Probability Models and Conditional Random Fields(1)--Naive Bayes
- [导入]Wiley - VERIFICATION AND VALIDATION FOR QUALITY OF UML 2.0 MODELS
- Dirichlet Process Gaussian Mixture Models
- Save and Load Your Keras Deep Learning Models