What is a single-linkage, a complete linkage and an average linkage? How do they influence a hierarchical agglomeration?
Can you give advantages/drawbacks of a hierarchical clustering vs. k-means clustering?
"When using k-means, the greatest issue is to know k". Can you explain this sentence? Do you know a way to discover k?
Given the two-dimensional objects the teacher will plot, what algorithm(s) (with what configuration) would you suggest to discover the two "obvious" clusters? Instead of directly processing these objects, what derived attribute seems to make more sense?
How do k-means, fuzzy c-means, and EM relate? What are their respective advantages and drawbacks?
Do you know a clustering algorithm that is sensitive to outliers? Why is it so? Other clustering algorithms can actually help in discovering outliers. Can you list two such algorithms?
Clustering algorithms can be seen as optimization techniques. What is optimized? How does that relate to assessing a clustering?
