Mean Shift Clustering Quiz

Answers and explanations appear as you take the quiz.

1. What direction does a mean-shift update move a seed?
   • Toward a local weighted mean of nearby observations
   • Toward the global mean of all observations at once
   • Toward the closest centroid from a separately fitted k-means model
   • Toward the single observation with the largest coordinate value
2. What determines how smooth or locally detailed the density modes are?
   • The stopping tolerance for seed movement
   • Bandwidth
   • The number of candidate seeds
   • The threshold for merging converged centers
3. If bandwidth is increased substantially, what may happen to nearby density peaks?
   • They become supervised labels
   • They must split into more modes
   • They may merge into a broader mode
   • Their feature units are removed
4. Why can a narrow bandwidth produce too many clusters?
   • It sets the number of clusters to one
   • It forces every seed to the global mean
   • It removes all local density variation
   • It may preserve small local bumps as separate modes
5. Why standardize features before distance-based mean shift when units differ greatly?
   • It changes the kernel to a different family automatically
   • A large-unit feature can dominate kernel neighborhoods
   • It makes the chosen bandwidth independent of feature units
   • It makes the inferred number of modes proportional to the feature scale
6. How is mean shift's cluster count determined?
   • By the number of input features
   • By a required k parameter in every implementation
   • By the density modes reached under chosen settings
   • By the count of observations at the origin
7. What can seed selection affect?
   • The probability normalization across classes
   • The meaning of bandwidth
   • Whether observations have feature values
   • Which starting locations are explored and computational cost
8. Which data condition can make mean shift less suitable?
   • Clusters with very different density scales
   • Several well-separated density peaks
   • Features with a meaningful common scale
   • A dataset where local modes define groups