Machine Learning Unsupervised Learning 2 — Questions and Answers
Question 1: Which metric is commonly used to evaluate the quality of clusters when ground truth labels are unavailable?
- Silhouette score (Correct answer)
- F1 score
- Precision
- Recall
Correct answer: Silhouette score
The silhouette score measures how similar a point is to its own cluster compared to other clusters, requiring no ground truth labels.
Question 2: What does the 'elbow method' help determine in K-Means clustering?
- The optimal number of clusters (Correct answer)
- The best distance metric
- The number of iterations needed
- The outlier threshold
Correct answer: The optimal number of clusters
The elbow method plots inertia vs. number of clusters and identifies the point where adding more clusters yields diminishing returns.
Question 3: Which dimensionality reduction technique is specifically designed to preserve local neighborhood structure?
- t-SNE (Correct answer)
- PCA
- SVD
- Factor Analysis
Correct answer: t-SNE
t-SNE (t-distributed Stochastic Neighbor Embedding) focuses on preserving local pairwise distances and is widely used for visualization.
Question 4: In Gaussian Mixture Models (GMM), what algorithm is used to estimate the parameters?
- Expectation-Maximization (EM) (Correct answer)
- Gradient Descent
- Backpropagation
- Newton-Raphson
Correct answer: Expectation-Maximization (EM)
The EM algorithm alternates between the E-step (computing soft cluster assignments) and M-step (updating parameters) to fit GMMs.
Question 5: What is the primary advantage of DBSCAN over K-Means clustering?
- It can find arbitrarily shaped clusters and identify outliers (Correct answer)
- It is faster on large datasets
- It requires fewer hyperparameters
- It always produces globally optimal clusters
Correct answer: It can find arbitrarily shaped clusters and identify outliers
DBSCAN identifies clusters of arbitrary shape based on density and automatically marks sparse points as noise/outliers.
Question 6: In PCA, what does each principal component represent?
- A direction of maximum variance in the data (Correct answer)
- A cluster centroid
- A latent class label
- A feature importance score
Correct answer: A direction of maximum variance in the data
Each principal component is an orthogonal direction in feature space along which the data has the most variance.
Question 7: Which of the following is a hard clustering method?
- K-Means (Correct answer)
- Gaussian Mixture Model
- Fuzzy C-Means
- Latent Dirichlet Allocation
Correct answer: K-Means
K-Means assigns each point to exactly one cluster (hard assignment), unlike soft/probabilistic methods that assign fractional membership.
Machine Learning Unsupervised Learning 2
Which metric is commonly used to evaluate the quality of clusters when ground truth labels are unavailable?