Clustering 2

Random Initialization

Random initialization

• Should have K < m
• Randomly pick K training examples
• Set μ1, …, μk equal to these K examples

Local optima

解法: initial 多次！　看看出來的結果是不是都大約一致~

Choosing the number of clusters

What is the right value of k?

阿到底要分幾個？　你知嗎～～～～

Chossing the value of K

運氣好找的到 Elbow (左圖) 明顯的折彎
運氣不好的話 難找阿 (右圖)

• EX:

Choosing the number of clusters