Notebooks

These executable notes work through small machine-learning and performance questions. The public repository validates every notebook file and runs the curated set from top to bottom in CI.

The rendered notebooks are hosted by the separate notebooks project. Each link below opens its output; this page keeps the collection connected to the engineering record.

A perceptron you can audit

A two-weight implementation with deterministic data, an update trace, and a held-out check.

• Read the rendered notebook
• View the source

NumPy versus a Python loop

Repeated trials, medians and IQRs across four vector sizes, with equivalence assertions and environment disclosure.

• Read the rendered notebook
• View the source

A logistic-regression baseline

A scaled, stratified baseline with an explicit class mapping and variation across five folds.

• Read the rendered notebook
• View the source

The survey notebook remains historical analysis and is under review while its population, skip logic, ordinal bins, and subgroup denominators are rebuilt.