mean-shift clustering

• goal: find dense area of points

• steps

• 1. start your sliding window (aka a sphere) at some point
• 2. then keep moving the centre of your window to the mean of the datapoints
• 3. keep doing this until it’s stopped
• 4. Do this multiple times to find all the clusters
• 5. remove all of the duplicate spheres (at the same point/near each other) at the end of the algo

• pros:

  • you don’t have to select the number of clusters

• con:

  • selecting the right radius is hard