Some references
- http://www.52nlp.cn/2012/04
- https://github.com/Computing-Intelligence/References
- http://web.stanford.edu/class/cs224n/
- https://study.163.com/course/courseLearn.htm?courseId=1004570029#/learn/video?lessonId=1052316982&courseId=1004570029
1 吴恩达ML
Clustering
![](https://img.haomeiwen.com/i8909247/0cdf4be70cba404c.png)
K-Means (usually two steps in each inner loop)
![](https://img.haomeiwen.com/i8909247/6b2962989f968f4f.png)
First step: Cluster assignments steps
Next: move cluster centroid to where close to the average of other dots
![](https://img.haomeiwen.com/i8909247/437a4e1490b2f70d.png)
Optimization objective
![](https://img.haomeiwen.com/i8909247/dffef959fa1b183a.png)
![](https://img.haomeiwen.com/i8909247/7b4110ee1ef0445d.png)
Random Initialization
![](https://img.haomeiwen.com/i8909247/0da2d6e32dc385bf.png)
![](https://img.haomeiwen.com/i8909247/99d3cbf762fd35a1.png)
![](https://img.haomeiwen.com/i8909247/dbb6dccba8f5ceec.png)
Choosing the number of K
更多的时候是自己设置,there isnt always a good answer
![](https://img.haomeiwen.com/i8909247/74712c4dc37c3a47.png)
![](https://img.haomeiwen.com/i8909247/49b68db17e0340ba.png)
2. 北理工ML
主题模型
![](https://img.haomeiwen.com/i8909247/66a1d629d8af1c00.png)
![](https://img.haomeiwen.com/i8909247/9184edfc0f6c8290.png)
MK矩阵:每个词w 在 m行文档K列主题 的分布矩阵
![](https://img.haomeiwen.com/i8909247/d2594c4b4b722092.png)
![](https://img.haomeiwen.com/i8909247/dd143625ae1a41bf.png)
3. 周志华ML
聚类
- 聚类试图将数据集中的样本划分为若干个通常是不相交的子集,每个子集 称为一个"簇" (cluster). 聚类过程仅能自动形成簇结构,簇所对应的概念语义需由使用者来把握和命名.
- 性能度量:聚类结果的"簇内相似 度" (intra-cluster similarity)高且"簇间相似度" (inter-cluster similarity)低. 聚类性能度量大致有两类. 一类是将聚类结果与某个"参考模 型" (reference model)进行比较,称为"外部指标" (external i丑dex); 另一 类是直接考察聚类结果而不利用任何参考模型,称为"内部指标" (internal index).
Prototype-based clustering
K-means
网友评论