Clustering and dimensionality reduction quiz

Question 1

What is the main goal of clustering in machine learning?

Accepted Answer

Group similar data points without using labels

Answer

Predict a numeric target from labeled data

Answer

Classify emails using known spam labels

Answer

Reduce overfitting in a decision tree

Question 2

Why is clustering considered unsupervised learning?

Accepted Answer

Because it finds patterns without labeled outputs

Answer

Because it uses labeled data to train

Answer

Because it requires no input data

Answer

Because it always uses neural networks

Question 3

In K-means clustering, what does the parameter K represent?

Accepted Answer

The number of clusters to form

Answer

The number of features in the dataset

Answer

The number of iterations to run

Answer

The distance metric used

Question 4

Describe the two main steps that K-means repeats in each iteration.

Accepted Answer

Assign points to nearest centroid, then recompute centroids

Answer

Sort points then delete outliers

Answer

Split clusters then merge them

Answer

Normalize data then label it

Question 5

What distance measure does standard K-means typically use, and why does scaling features matter?

Accepted Answer

Euclidean distance; scaling prevents large-range features from dominating

Answer

Cosine distance; scaling is irrelevant

Answer

Manhattan distance; scaling increases the number of clusters

Answer

Hamming distance; scaling removes outliers

Question 6

What is the elbow method used for in clustering?

Accepted Answer

To choose the number of clusters K

Answer

To measure classification accuracy

Answer

To select the learning rate

Answer

To remove correlated features

Question 7

What does 'inertia' (within-cluster sum of squares) measure in K-means?

Accepted Answer

The sum of squared distances from points to their centroids

Answer

The number of points per cluster

Answer

The accuracy of the cluster labels

Answer

The total number of iterations

Question 8

Why does K-means sometimes give different results on different runs, and how is this addressed?

Accepted Answer

Because of random initial centroids; fix by running multiple inits and keeping the best

Answer

Because it shuffles labels; fix by adding labels

Answer

Because of feature scaling; fix by removing scaling

Answer

Because K changes each run; fix by lowering K

Question 9

What is k-means++ and what problem does it solve?

Accepted Answer

A smarter centroid initialization that spreads starting centers apart

Answer

A method to pick K automatically

Answer

A way to compute distances faster

Answer

A technique to label clusters with names

Question 10

How does hierarchical agglomerative clustering build clusters?

Accepted Answer

By starting with each point alone and merging closest clusters

Answer

By starting with one cluster and splitting it repeatedly

Answer

By assigning points to K fixed centroids

Answer

By projecting data onto principal components

Question 11

What is a dendrogram and how do you read it?

Accepted Answer

A tree showing the order and distance of cluster merges

Answer

A scatter plot of two principal components

Answer

A bar chart of cluster sizes

Answer

A heatmap of feature correlations

Question 12

In hierarchical clustering, what does the choice of 'linkage' control?

Accepted Answer

How the distance between two clusters is computed

Answer

The number of clusters to keep

Answer

The learning rate of the algorithm

Answer

The number of features used

Question 13

What is one key advantage of hierarchical clustering over K-means?

Accepted Answer

It does not require specifying the number of clusters beforehand

Answer

It is always faster than K-means

Answer

It guarantees the global optimum

Answer

It needs no distance measure

Question 14

What kind of clusters can DBSCAN find that K-means struggles with?

Accepted Answer

Arbitrary-shaped, density-based clusters with noise detection

Answer

Only perfectly circular clusters

Answer

Clusters that require labels

Answer

Clusters defined only by their centroids

Question 15

In DBSCAN, what do the parameters eps and min_samples control?

Accepted Answer

The neighborhood radius and minimum points for a dense region

Answer

The number of clusters and iterations

Answer

The learning rate and momentum

Answer

The number of principal components and variance

Question 16

What is dimensionality reduction and why is it useful?

Accepted Answer

Reducing the number of features while keeping key information

Answer

Adding more features to improve accuracy

Answer

Increasing the number of data points

Answer

Removing all but one data sample

Question 17

What is the 'curse of dimensionality'?

Accepted Answer

Problems from data sparsity and weak distances in high dimensions

Answer

A bug that occurs only in 2D data

Answer

A guarantee that more features always help

Answer

An error caused by too few features

Question 18

What does Principal Component Analysis (PCA) do?

Accepted Answer

Finds orthogonal directions of maximum variance and keeps the top ones

Answer

Randomly drops half the features

Answer

Groups points into clusters

Answer

Adds polynomial features to the data

Question 19

In PCA, what does the first principal component represent?

Accepted Answer

The direction of maximum variance in the data

Answer

The least important feature

Answer

The average of all features

Answer

The cluster centroid

Question 20

What does 'explained variance ratio' tell you in PCA?

Accepted Answer

The fraction of total variance captured by each component

Answer

The accuracy of a classifier

Answer

The number of clusters found

Answer

The distance between two points

Question 21

Why is it important to standardize features before applying PCA?

Accepted Answer

So that high-variance large-scale features do not dominate the components

Answer

To increase the number of features

Answer

Because PCA cannot run on numeric data

Answer

To convert data into labels

Question 22

Is PCA a linear or non-linear dimensionality reduction technique, and what is one consequence?

Accepted Answer

Linear; it may miss non-linear structure in the data

Answer

Non-linear; it captures any curved structure

Answer

Linear; it always captures non-linear patterns

Answer

Non-linear; it requires labels

Question 23

What is t-SNE primarily used for?

Accepted Answer

Visualizing high-dimensional data by preserving local structure

Answer

Predicting continuous targets

Answer

Encrypting data

Answer

Increasing dataset size

Question 24

Why should you be cautious interpreting distances between clusters in a t-SNE plot?

Accepted Answer

Because t-SNE preserves local structure, so inter-cluster distances are unreliable

Answer

Because t-SNE preserves only global distances

Answer

Because t-SNE adds labels automatically

Answer

Because t-SNE always gives identical plots

Question 25

What is the silhouette score and what range does it take?

Accepted Answer

-1 to +1, higher means better-defined clusters

Answer

0 to 100, higher is better

Answer

0 to 1 only, lower is better

Answer

Any real number with no bound

Question 26

What is the difference between a centroid and a medoid in clustering?

Accepted Answer

A centroid is the mean point; a medoid is an actual representative data point

Answer

They are identical terms

Answer

A medoid is always the mean; a centroid is a real point

Answer

Both must be real data points

Question 27

How does the number of features relate to how PCA can reduce dimensions?

Accepted Answer

PCA keeps the top variance components, fewer than the original features

Answer

PCA can create more components than original features

Answer

PCA always keeps exactly one component

Answer

PCA increases the number of features

Question 28

What is feature scaling's general effect on distance-based clustering algorithms?

Accepted Answer

It ensures all features contribute fairly to distances

Answer

It has no effect on clustering

Answer

It always reduces the number of clusters

Answer

It converts the task into supervised learning

Question 29

Can K-means handle clusters of very different sizes and densities well? Why or why not?

Accepted Answer

No, it favors equal-sized spherical clusters

Answer

Yes, it handles all shapes and sizes perfectly

Answer

Yes, because it uses density estimates

Answer

No, because it requires labeled data

Question 30

What does it mean that principal components are orthogonal?

Accepted Answer

They are perpendicular and mutually uncorrelated

Answer

They are parallel and highly correlated

Answer

They are identical copies of the features

Answer

They always point along original axes

Test yourself