Machine Learning Clustering 4 — Questions and Answers
Question 1: In a Gaussian Mixture Model, what does each component's covariance matrix control?
- The number of data points assigned to the component
- The shape and orientation of the cluster (Correct answer)
- The mixing proportion of the component
- The mean position of the cluster
Correct answer: The shape and orientation of the cluster
The covariance matrix determines whether a Gaussian component is spherical, elongated, or rotated in any direction.
Question 2: Which cluster validity index penalizes clusters that are far from each other and rewards clusters with low intra-cluster scatter?
- Silhouette Score
- Davies-Bouldin Index
- Calinski-Harabasz Index (Correct answer)
- Rand Index
Correct answer: Calinski-Harabasz Index
The Calinski-Harabasz Index is the ratio of between-cluster dispersion to within-cluster dispersion; higher values indicate better-defined clusters.
Question 3: What does the 'elbow method' help determine in K-Means clustering?
- The best linkage criterion for hierarchical clustering
- The optimal epsilon value for DBSCAN
- The appropriate number of clusters K (Correct answer)
- The best distance metric to use
Correct answer: The appropriate number of clusters K
The elbow method plots inertia vs. K and looks for the point where adding more clusters yields diminishing returns.
Question 4: Which type of hierarchical clustering starts with all points in one cluster and recursively splits them?
- Agglomerative clustering
- Divisive clustering (Correct answer)
- Spectral clustering
- Ward clustering
Correct answer: Divisive clustering
Divisive hierarchical clustering uses a top-down approach, splitting the single large cluster into smaller ones at each step.
Question 5: What is the role of the minPts parameter in DBSCAN?
- It sets the maximum number of clusters
- It defines the minimum number of points required to form a dense region (Correct answer)
- It controls the learning rate
- It sets the distance threshold for merging
Correct answer: It defines the minimum number of points required to form a dense region
A point is a core point in DBSCAN if at least minPts points (including itself) fall within its epsilon-radius neighborhood.
Question 6: Which preprocessing step most commonly improves K-Means performance on real-world data?
- One-hot encoding all features
- Feature scaling (standardization or normalization) (Correct answer)
- Increasing the number of iterations
- Removing all outliers before clustering
Correct answer: Feature scaling (standardization or normalization)
K-Means uses Euclidean distance, so features on different scales can dominate; standardizing ensures all features contribute equally.
Question 7: In spectral clustering, which step transforms the data before applying K-Means?
- PCA dimensionality reduction
- Mapping data into eigenvector space of the graph Laplacian (Correct answer)
- Computing pairwise cosine similarities
- Normalizing all features to [0, 1]
Correct answer: Mapping data into eigenvector space of the graph Laplacian
Spectral clustering constructs an affinity graph, computes the graph Laplacian, and clusters in the space of its top eigenvectors.
Machine Learning Clustering 4
In a Gaussian Mixture Model, what does each component's covariance matrix control?