Week 10: Clustering

Question 1

Clustering

Answer 1

A method of grouping samples based on how similar they are to each other.

Question 2

K-Means Clustering

Answer 2

A method of clustering where samples are grouped based on the closest centroid. Centroids are recalculated after each iteration by averaging the coordinates of all samples assigned to the centroid.

Question 3

Fuzzy K-means Clustering

Answer 3

A method assigning a probability of each sample belonging to each centroid. Centroids and probabilities are updated in each iteration.

Question 4

Iterative Clustering

Answer 4

A subset of clustering that focuses on local optimisation over global optimisation.

Question 5

Basic Iterative Min-Squared-Error Clustering

Answer 5

It’s a type of iterative clustering that’s essentially K-means clustering done sequential for each sample.

Question 6

Hierarchical Clustering

Answer 6

A type of clustering where the samples are split into subclusters that have a hierarchy.

Question 7

Agglomerative Hierarchical Clustering

Answer 7

A type of hierarchical clustering that begins with each sample in its own cluster. Clusters are merged together until only one cluster remains.

Question 8

Competitive Learning

Answer 8

Similar to K-means, except it’s done sequentially and only the assigned centroid is updated each step.

Question 9

Basic-leader-follower Clustering

Answer 9

It’s done sequentially, with the first sample being the first centroid. If the distance to the centroid with within a threshold, the centroid is kept and updated. If the distance isn’t within the threshold, a new centroid is created with the current sample.