MS-DS Master of Data science MS-DS Master of Data science Unsupervised Learning Techniques Questions and Answers 3

Question 1

What advantage does DBSCAN have over K-Means for spatial data with irregular cluster shapes?

Accepted Answer

DBSCAN can discover clusters of arbitrary shape based on density

Answer

DBSCAN always runs faster than K-Means

Answer

DBSCAN requires fewer hyperparameters than K-Means

Answer

DBSCAN guarantees globally optimal cluster assignments

Question 2

In K-Means clustering, what problem can arise from poor centroid initialization?

Accepted Answer

Convergence to a suboptimal local minimum

Answer

The algorithm fails to terminate

Answer

Clusters are always perfectly balanced

Answer

The number of clusters changes automatically

Question 3

Which linkage criterion in hierarchical clustering measures the distance between the closest members of two clusters?

Accepted Answer

Single linkage

Answer

Complete linkage

Answer

Average linkage

Answer

Ward's linkage

Question 4

What is the role of the covariance matrix in a Gaussian Mixture Model?

Accepted Answer

It captures the shape, orientation, and spread of each Gaussian component

Answer

It determines the number of clusters

Answer

It sets the learning rate for parameter updates

Answer

It normalizes input features before clustering

Question 5

What does the reconstruction error of an autoencoder indicate about the input data?

Accepted Answer

How well the compressed representation captures the essential features of the input

Answer

The classification accuracy of the model

Answer

The number of layers needed in the network

Answer

The optimal learning rate for training

Question 6

Which method helps determine the optimal number of clusters for K-Means by plotting within-cluster sum of squares against cluster count?

Accepted Answer

The Elbow Method

Answer

The Silhouette Method

Answer

Cross-validation

Answer

The Gap Statistic