Gaussian Mixture Models in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a Gaussian Mixture Model (GMM)?

A Gaussian Mixture Model is a way to represent data as a mix of several bell-shaped curves (Gaussians). Each curve represents a group or cluster in the data.

intermediate

How does GMM differ from K-Means clustering?

GMM assumes data points come from a mix of Gaussian distributions and can assign probabilities to clusters, while K-Means assigns each point to exactly one cluster without probabilities.

intermediate

What is the role of the Expectation-Maximization (EM) algorithm in GMM?

EM helps find the best parameters for the Gaussian curves by repeating two steps: guessing which points belong to which curve (Expectation), then updating the curves to better fit those points (Maximization).

beginner

What are the main parameters of a Gaussian component in GMM?

Each Gaussian has a mean (center), covariance (shape and spread), and a weight (how much it contributes to the overall mix).

beginner

Why is GMM considered a soft clustering method?

Because it assigns probabilities to each data point for belonging to each cluster, instead of a hard yes/no assignment.

What does each component in a Gaussian Mixture Model represent?

A A decision tree node

B A Gaussian distribution representing a cluster

C A single data point

D A linear regression line

Each component is a Gaussian distribution that models one cluster in the data.

Which algorithm is commonly used to estimate parameters in GMM?

A Expectation-Maximization

B Gradient Descent

C K-Nearest Neighbors

D Support Vector Machine

Expectation-Maximization (EM) is used to find the best parameters for the Gaussian components.

What does the covariance matrix in a Gaussian component describe?

A The center of the cluster

B The probability of the cluster

C The number of clusters

D The shape and spread of the cluster

Covariance describes how data spreads and the shape of the Gaussian curve.

In GMM, what does a higher weight for a Gaussian component mean?

A It contributes less to the overall model

B It has fewer data points

C It contributes more to the overall model

D It has a smaller spread

A higher weight means the component has more influence in the mixture.

Why might GMM be preferred over K-Means for clustering?

A GMM can model clusters with different shapes and sizes

B GMM is faster to compute

C GMM only works with binary data

D GMM does not require parameter tuning

GMM can model clusters with different shapes because it uses covariance matrices, unlike K-Means which assumes spherical clusters.

Explain how the Expectation-Maximization algorithm works in Gaussian Mixture Models.

Think about guessing cluster membership and then improving the guess.

Describe the difference between hard clustering and soft clustering with examples.

Consider how certain or uncertain the cluster assignment is.