t-SNE for visualization in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is t-SNE used for in machine learning?

t-SNE is used to reduce high-dimensional data to two or three dimensions to help visualize complex data patterns in a simple, easy-to-understand way.

intermediate

How does t-SNE preserve data structure during visualization?

t-SNE preserves local relationships by keeping similar points close together in the low-dimensional space, making clusters visible while allowing some distortion of global structure.

intermediate

What is the main difference between PCA and t-SNE?

PCA is a linear method that preserves global variance, while t-SNE is a nonlinear method focused on preserving local similarities for better cluster visualization.

advanced

What role does the 'perplexity' parameter play in t-SNE?

Perplexity controls the balance between local and global aspects of the data; it roughly sets how many neighbors each point influences during the mapping.

advanced

Why should t-SNE results be interpreted carefully?

Because t-SNE can distort global distances and is sensitive to parameters, its plots show clusters well but should not be used to infer exact distances or relationships between clusters.

What is the primary goal of t-SNE?

A Increase the number of features in data

B Visualize high-dimensional data in 2D or 3D

C Train a classification model

D Generate synthetic data

t-SNE reduces data dimensions to 2 or 3 to help visualize complex data.

Which aspect does t-SNE focus on preserving?

A Local similarities between nearby points

B Global distances between all points

C Exact numeric values of features

D Data labels

t-SNE preserves local similarities to show clusters clearly.

What does the 'perplexity' parameter affect in t-SNE?

A Number of output dimensions

B Number of iterations

C Learning rate speed

D Balance between local and global data structure

Perplexity controls how many neighbors influence each point, balancing local/global structure.

Why is t-SNE not suitable for very large datasets without adjustments?

A It is computationally expensive and slow

B It increases data dimensionality

C It only works with images

D It requires labeled data

t-SNE can be slow and costly on large datasets without special techniques.

Which method is a linear alternative to t-SNE?

A Random Forest

B K-means

C PCA

D Neural Networks

PCA is a linear dimensionality reduction method often compared to t-SNE.

Explain how t-SNE helps visualize complex data and what makes it different from other dimensionality reduction methods.

Think about how t-SNE shows clusters clearly by focusing on neighbors.

Describe the importance of the 'perplexity' parameter in t-SNE and how changing it might affect the visualization.

Consider how many neighbors each point 'sees' during mapping.