Why One-Hot Encoding Turned 25MB Into 1.4GB
Here's a real number, produced by running the code below, not borrowed from anywhere: a single categorical column, 500,000 rows, 3,000 unique values, sitting at a modest 25.6MB in memory. One-hot enco
Search for a command to run...
Articles tagged with #technical-writing-1
Here's a real number, produced by running the code below, not borrowed from anywhere: a single categorical column, 500,000 rows, 3,000 unique values, sitting at a modest 25.6MB in memory. One-hot enco
It's a familiar shape of incident, and versions of it show up constantly in backend post-mortems: everything runs fine in staging against a few thousand rows, ships to production, and then real traffi
For years, pandas code like this had a reputation for silently not doing what it looked like it was doing: df[df["score"] > 90]["grade"] = "A" Sometimes it worked. Sometimes it didn't — and pandas w
Databases have a classic trick for finding matching rows fast: build a hash table once, then look things up in it instead of scanning everything from scratch every time you need an answer. It's one of
Ask most people to picture a home network under attack, and they imagine the same scene: someone cracking a Wi-Fi password, slipping past the router, and rifling through a laptop. So the advice we've
Normalizing rescales your data onto a common scale — which matters a lot for machine learning models that are sensitive to the raw magnitude of features, and matters even for something as simple as co