Lecture 14 - Expectation-Maximization Algorithms | Stanford CS229 Machine Learning

Lecture 14 - Expectation-Maximization Algorithms | Stanford CS229: Machine Learning (Autumn 2018)

• for k-means clustering, Prof. Ng picks the k according to the goal of the clustering and relevant reasoning based on domain knowledge rather than using AIC or BIC. E.g. the market has bandwidth for 4 for market campaigns rather than 100, so choose k=4 is in market segmentation.
• 47:22 “why it’s a maximum likelihood estimation algorithm and why we can expect it to converge”