juniorDimensionality Reduction
What is PCA in dimensionality reduction?
Updated May 16, 2026
Short answer
PCA is a technique that transforms data into principal components that capture maximum variance.
Deep explanation
PCA finds orthogonal axes (principal components) that maximize variance. It uses eigen decomposition of covariance matrix or SVD to project data into fewer dimensions while preserving information.
Real-world example
Used in facial recognition to compress image data into eigenfaces.
Common mistakes
- Thinking PCA selects original features instead of creating new ones.
Follow-up questions
- What does each principal component represent?
- Is PCA supervised or unsupervised?