March 15, 2026
Old classic, new comment war
A Visual Introduction to Machine Learning
Beloved throwback explainer returns—fans cheer, skeptics ask where the rest went
TLDR: A 2015 visual guide to machine learning is back, charming many with its simple house-price example while others gripe that the page cuts off and feels dated. Fans shout “masterpiece,” skeptics want modern explainers like Transformers, and many point beginners to StatQuest—proof people crave clear, complete AI explainers.
An old-school crowd-pleaser just resurfaced: a 2015 visual guide that teaches machine learning by sorting homes in San Francisco vs. New York using hills, price-per-square-foot, and a simple “choose-this-or-that” flowchart called a decision tree. The vibes? Nostalgia meets nitpicking. One camp is swooning, calling it a masterpiece and saying it was ahead of its time. Another camp storms in with, “Wait… where’s the rest of it?” because the page literally cuts off mid-word—and yes, that cliffhanger sent the comment section into detective mode.
The fans are sharing more candy for the brain: a stash of interactive explainers via this list, while the pragmatists are out here handing you study hacks—cue the chorus plugging Josh Starmer on YouTube and StatQuest, hailed as the easiest path for beginners. Meanwhile, the futurists crashed the party asking for an r2d3-style tour of Transformers’ attention (that’s the “focus trick” behind modern chatbots), proving the bar for visual explainers is sky-high in 2026.
And yes, the community had jokes: the magical “$1776 per square foot” line got the most side-eye—“patriotic pricing,” anyone? The debate boils down to this: timeless starter guide vs. dated, incomplete throwback. Either way, the comments turned a gentle explainer into a full-blown nostalgia brawl—classic internet energy.
Key Points
- •The article explains classification using a housing dataset to distinguish San Francisco and New York homes.
- •Initial intuition uses elevation (around >240 ft) to identify San Francisco homes; adding price per square foot (> $1776) improves separation.
- •A seven-dimensional dataset is explored with a scatterplot matrix, showing patterns not easily separated by simple visual rules.
- •Decision trees are introduced to find data-driven split points, with tradeoffs leading to false positives and false negatives.
- •Training proceeds recursively: after finding a best split, the process repeats on subsets to create additional splits and purer branches.