Questions

1. Is a half-moon-shaped dataset a convex cluster?
2. A bidimensional dataset is made up of two half-moons. The second one is fully contained in the concavity of the first one. Which kind of kernel can easily allow the separation of the two clusters (using spectral clustering)?
3. After applying the DBSCAN algorithm with ε=1.0, we discover that there are too many noisy points. What should we expect with ε=0.1?
4. K-medoids is based on the Euclidean metric. Is this correct?
5. DBSCAN is very sensitive to the geometry of the dataset. Is this correct?
6. A dataset contains 10,000,000 samples and can be easily clustered using a large machine using K-means. Can we, instead, use a smaller machine and mini-batch K-means?
7. A cluster has a standard deviation equal to 1.0. After applying a noise N(0, 0.005), 80% of the original assignments are changed. Can we say that such a cluster...