1. Clustering Unsupervised learning introduction Machine Learning

2. Andrew Ng Supervised learning Training set:

3. Andrew Ng Unsupervised learning Training set:

4. Andrew Ng Applications of clustering Organize computing clusters Social network analysis Image credit: NASA/JPL-Caltech/E. Churchwell (Univ. of Wisconsin, Madison) Astronomical data analysis Market segmentation

5. Clustering K-means algorithm Machine Learning

6. Andrew Ng

15. Input: - (number of clusters) -Training set (drop convention) K-means algorithm

16. Andrew Ng Randomly initialize cluster centroids K-means algorithm Repeat { for = 1 to := index (from 1 to ) of cluster centroid closest to for = 1 to := average (mean) of points assigned to cluster }

17. Andrew Ng K-means for non-separated clusters T-shirt sizing Height Weight

18. Clustering Optimization objective Machine Learning

19. Andrew Ng K-means optimization objective = index of cluster (1,2,…, ) to which example is currently assigned = cluster centroid ( ) = cluster centroid of cluster to which example has been assigned Optimization objective:

20. Andrew Ng Randomly initialize cluster centroids K-means algorithm Repeat { for = 1 to := index (from 1 to ) of cluster centroid closest to for = 1 to := average (mean) of points assigned to cluster }

21. Clustering Random initialization Machine Learning

22. Andrew Ng Randomly initialize cluster centroids K-means algorithm Repeat { for = 1 to := index (from 1 to ) of cluster centroid closest to for = 1 to := average (mean) of points assigned to cluster }

23. Andrew Ng Random initialization Should have Randomly pick training examples. Set equal to these examples.

24. Andrew Ng Local optima

25. Andrew Ng For i = 1 to 100 { Randomly initialize K-means. Run K-means. Get. Compute cost function (distortion) } Pick clustering that gave lowest cost Random initialization

26. Clustering Choosing the number of clusters Machine Learning

27. Andrew Ng What is the right value of K?

28. Andrew Ng Choosing the value of K Elbow method: Cost function (no. of clusters) Cost function (no. of clusters)

29. Andrew Ng Choosing the value of K Sometimes, you’re running K-means to get clusters to use for some later/downstream purpose. Evaluate K-means based on a metric for how well it performs for that later purpose. E.g. T-shirt sizing Height Weight T-shirt sizing Height Weight