Principal Components as New Basis
What you'll learn: How principal components form a new coordinate system that better represents your data's structure.
The Core Idea
After performing eigendecomposition on the covariance matrix, you've extracted eigenvectors (directions) and eigenvalues (importance). These eigenvectors are your principal components, and they function as a completely new set of axes—a new basis—for describing your data.
Think of it like rotating a map. Imagine you have a street map where all the roads run at 45-degree angles to the page edges. Reading coordinates is awkward! But if you rotate the map so the main road aligns with the page's vertical axis, suddenly positions become much easier to describe. That's what principal components do: they rotate your coordinate system to align with the data's natural variance structure.
Why This Matters
In your original coordinate system, data might be scattered messily across multiple features. The principal components create orthogonal axes (perpendicular to each other) where:
- The first principal component points in the direction of maximum variance
- The second points in the direction of maximum remaining variance (perpendicular to the first)
- And so on...
Each data point that was originally described by coordinates like (x₁, x₂, ..., xₙ) can now be described using coordinates in this new basis—essentially projecting your data onto these new axes.
The beauty? Most of the information (variance) concentrates in the first few principal components, letting you drop the rest without losing much information.
Key Takeaway: Principal components are orthogonal axes forming a new coordinate system aligned with data variance—they're not just directions, but a complete new basis for representing your data more efficiently.