Machine Learning Unsupervised Learning 3 — Questions and Answers
Question 1: What is an autoencoder in the context of unsupervised learning?
- A neural network trained to reconstruct its input through a bottleneck (Correct answer)
- A classifier that labels unlabeled data
- A generative model that creates synthetic labels
- A clustering algorithm based on neural networks
Correct answer: A neural network trained to reconstruct its input through a bottleneck
An autoencoder compresses input to a lower-dimensional latent code (encoder) and then reconstructs the original input (decoder).
Question 2: Which linkage criterion in hierarchical clustering tends to produce compact, spherical clusters?
- Ward linkage (Correct answer)
- Complete linkage
- Single linkage
- Average linkage
Correct answer: Ward linkage
Ward linkage minimizes the total within-cluster variance at each merge step, producing compact and roughly spherical clusters.
Question 3: What problem does ICA (Independent Component Analysis) solve that PCA does not?
- Separating statistically independent non-Gaussian source signals (Correct answer)
- Reducing dimensionality while maximizing variance
- Finding the optimal number of clusters
- Removing correlated features
Correct answer: Separating statistically independent non-Gaussian source signals
ICA finds a decomposition into statistically independent components, making it suitable for blind source separation tasks like separating audio signals.
Question 4: What is a dendrogram used for in clustering?
- Visualizing the hierarchy of cluster merges in agglomerative clustering (Correct answer)
- Plotting the silhouette scores for each cluster
- Showing the decision boundary of a classifier
- Displaying the variance explained by principal components
Correct answer: Visualizing the hierarchy of cluster merges in agglomerative clustering
A dendrogram is a tree diagram showing the sequence of merges in hierarchical clustering, helping choose the number of clusters by cutting at a desired height.
Question 5: Which assumption does K-Means make about cluster shapes?
- Clusters are convex and isotropic (spherical) (Correct answer)
- Clusters can be any arbitrary shape
- Clusters follow a Gaussian distribution
- Clusters have equal density
Correct answer: Clusters are convex and isotropic (spherical)
K-Means uses Euclidean distance to assign points to the nearest centroid, implicitly assuming clusters are convex and roughly spherical.
Question 6: In topic modeling with Latent Dirichlet Allocation (LDA), what are documents modeled as?
- Mixtures of topics, where each topic is a distribution over words (Correct answer)
- Clusters of similar sentences
- Sequences of named entities
- Bag-of-words frequency vectors only
Correct answer: Mixtures of topics, where each topic is a distribution over words
LDA models each document as a mixture of latent topics, and each topic as a probability distribution over vocabulary words.
Question 7: What does 'inertia' measure in K-Means clustering?
- The sum of squared distances from each point to its assigned cluster centroid (Correct answer)
- The number of iterations until convergence
- The silhouette score of the clustering
- The distance between cluster centroids
Correct answer: The sum of squared distances from each point to its assigned cluster centroid
Inertia (also called within-cluster sum of squares, WCSS) measures how tightly packed the clusters are; lower inertia means more compact clusters.
Machine Learning Unsupervised Learning 3
What is an autoencoder in the context of unsupervised learning?