Machine Learning for Fun and Profit
summary
A real project for a state program has ~300 rule rows expressed as Cucumber tables and the workflow moves from hand‑coded ifs to learning and back: ID3 (via the AI4R gem) builds decision trees and AI4R outputs Ruby code for evaluation, but ID3 biases on continuous values unless discretized; the decision_tree gem implements C4.5‑style split‑point search and decision_tree handles continuous attributes (it chose a 2003.5 split), and a tiny custom decision_table gem matches literal rule tables; unit tests prove replacements equivalent and save days of work.
AI-written summary · about
chapters (5)
- 0:00Start
- 4:11TURNS INTO SIMPLE CODE
- 20:59ENTROPY BIAS
- 33:58LESSONS LEARNED . Don't write the code yourself!
- 35:41Resources
related talks
51mMachine Learning with Ruby
28mMachine Learning for Fun and Profit
35mIntelligent Ruby: Getting Started with Machine Learning
27mFive machine learning techniques that you can use in your Ruby apps today By Benjamin Curtis Machine learning is everywhere these days
27mI know Kung Fu! RubyGems for AI
39mRuleby: The Rules Engine for Ruby
01h 00mThe Promise and Limitations of AI
31mDefying Gravity: Teaching AI to Write Better Ruby