In partnership with
Become An AI Expert In Just 5 Minutes
If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.
This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.
Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.
Last week, K-Means Clustering strutted on stage, confidently placing everyone into neat little groups like a bouncer at an exclusive club.
Today’s contestant’s got different energy.
They’re not here to separate people into groups. They’re here to simplify chaos. Please welcome: Principal Component Analysis, aka PCA, the algorithm that walks into a high-dimensional dataset and says: “Relax. I can summarize this entire thing.”
📈 PCA- “The Dimensionality Magician”
Datasets can be, well…complicated. And messy. Taking one look at datasets with tons of features can quickly become overwhelming for us humans.
But for PCA? Piece of cake.
PCA takes one look at a complicated dataset and rewrites it in terms of just a few “main directions,” so it’s easier to visualize, compress, and model.
Principal Component Analysis (PCA) is a linear dimensionality reduction method that reduces the number of features in a dataset by creating uncorrelated variables called “principal components”.
In plain English, it takes a messy dataset with lots of columns and replaces it with a few new features that keep most of the important differences in the data. (hence “dimensionality reduction”).
🧠 Play By Play: How PCA Works
PCA re-expresses your data using new features that capture the biggest patterns of variation, using fewer dimensions.
Start with your data matrix: You have points scattered in many dimensions (the picture shows 2D, but it’s usually more). The cloud has a clear “main direction” where the points stretch out the most.
PCA finds the best new axes (middle): PCA doesn’t move points; it rotates the coordinate system.
PC1 = the direction the data spreads out the most (maximum variance).
PC2 = the next best direction for spread, but it must be perpendicular (orthogonal) to PC1.
Reduce dimensions by keeping only the important axis: To compress the data, you project each point onto PC1. Each point becomes a single number: “how far along PC1 it is.” You drop PC2, because it captured less of the important spread. That’s why the right plot above looks like points mostly along a line.
You kept PC1 and removed the less informative direction.
The result? You get simpler data (fewer features) while keeping most of the main patterns in the original dataset.
TLDR: The K-Means Mood Board
Step | What PCA Does | Vibe |
|---|---|---|
1 | Re-center the data (start from the middle) | “Reset.” |
2 | Find the main direction the data spreads (PC1). | “Biggest patterns.” |
3 | Finds the next direction at a right angle (PC2, etc.) | “Next biggest pattern.” |
4 | Keep the top few directions and drop the rest. | “Keep signal, cut noise.” |
5 | Rewrite each point using those new directions. | “Same data, simpler view.” |
To recap, PCA finds the direction in which your data varies the most. It rotates the view to line up with those directions, then keeps the strongest ones and drops the rest. Same data, just a simpler representation.
Conclusion
PCA is ultimately a clarity tool: it finds the few directions where your data changes the most, then lets you describe everything using those directions, instead of leaving you with a messy pile of original features.
PCA’s talent is helping you see structure, compress information, and make downstream modeling easier.
While PCA was the disciplined pro who cleaned up the stage with straight lines and explained variance, our next contestant is like an avant-garde performer who doesn’t care about straight lines…only who should be close to whom: t-SNE.
Stay tuned, as they make their grand entrance in the next episode!
