PCA vs t-SNE: PCA preserves global variance linearly, t-SNE preserves local structure nonlinearly

How can we teach computers to understand what we like?

Imagine you're shopping for a new coffee machine. You want one that makes coffee just right, not too bitter or too bland.

Think of the coffee machine as a computer learning from your taste. We want it to get better at making coffee you love over time.

Example

You give feedback on different coffee machines, telling them which ones taste good and which ones don't.

Remember this

It's like teaching the machine to recognize patterns in your preferences, so it can make better coffee choices.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

t-SNE preserves local structure

Can we see the hidden patterns in a cloud of data points?

Dimensionality reduction

Dimensionality reduction transforms high-dimensional data into low-dimensional space while preserving meaningful properties

Principal component analysis

Eigenvectors point along maximum variance

Eigenvalues and eigenvectors

Eigenvectors are unchanged in direction by a linear transformation

UMAP is faster than t-SNE

UMAP is faster due to approximate nearest neighbors and cross-entropy optimization

LoRA (machine learning)

LoRA uses r << d for efficient adaptation

Swipe through 100 ML concepts daily