Machine Learning notes — changelog
In the interest of making changes/updates as transparent and easy to follow as possible, I figured having a changelog here would make life a lot easier for everyone (and especially for me!).
Corrections by date
| Date | Chapter | Original | Updated |
|---|---|---|---|
| Jan 9th 2012 | 10 | if d is too small → this probably corresponds to a high variance problem | if d is too large → this probably corresponds to a high variance problem |
| Jan 9th 2012 | 10 | For high variance, we find the cross validation is low but training error is high | For high variance, we find the cross validation error is high but training error is low |
| Jan 9th 2012 | 1 | This is an example of a clarification problem | This is an example of a classification problem |
| Jan 30th 2012 | 8 | Define some additional terms z12 = g(...) | Define some additional terms z12 = ... |
| March 15th 2012 | 9 | i = the example index | i = the error of the affected node in the target layer |
2026 rewrite
The whole set was rebuilt from the original Evernote exports into clean, semantic HTML. The wording of the notes is unchanged apart from a short list of typo and notation fixes. Both the structural changes and every content fix are recorded in full in CHANGELOG.md, alongside these files.
The headline points:
- Every page is now valid HTML5 with real headings, properly nested lists, figures and code blocks — no
<font>tags, no inline styling, no unclosed elements. - Chapter 10 is included again, and the 167 diagrams that were missing from the source folders have been restored, so all 383 images resolve locally.
- Each chapter carries an "on this page" contents list and previous/next navigation.
- The pages are readable on a phone, in dark mode, and in print.