# Aabha Applied Machine Learning

Extract the ZIP, open `index.html`, and begin with the syllabus. The offline guide is a reading edition, not a Python runner. Keep the folder together so diagrams and links resolve. Work through Lessons 1–12 in order, run the corresponding examples, attempt practice, and only then open `solutions.html`.

This course builds on Aabha's Python for AI and Data, SQL for Data Analysis and AI, and Statistics with Data Storytelling. It teaches defensible supervised prediction using fictional depot jobs. It is a bridge into the separate Deep Learning Foundations course, which teaches tensors, gradients, neural-network training and a small CNN.

Python 3.12 is the tested practice route. See `SETUP.md` for exact commands and dependency pins. No GPU, account, API key or paid service is needed. Reading works offline after extraction; initial dependency installation and external references require internet.

- `SYLLABUS.md`: outcomes, sequence, prerequisites and estimated workload
- `lessons/` and `COURSEBOOK.md`: editable original teaching narrative
- `examples/`: ten progressively explained runnable demonstrations
- `exercises/`: learner-owned function and capstone starters, decision record
- `check_learner.py`: checks your code, not the reference answer
- `SOLUTIONS.md`, `solutions/`, `solutions.html`: separate answers
- `data/`: four synthetic CSV partitions; `DATA_DICTIONARY.md` explains provenance
- `courseml/`: inspectable generation, preprocessing, model comparison and evaluation code
- `outputs/`: reproducible reference results; final-test outputs are spoilers
- `tests/`: reference invariants and arithmetic tests, not a learner grade
- `assets/`: original instructional SVGs, with text descriptions
- `VALIDATION.md`: exact verified scope and outstanding limitations
- `integration/`: future integration handoff only; nothing has been published

The supplied reference freezes choices before its future-test run. As with any transparent teaching package, test labels and reference answers remain accessible. Treat the first test reveal as an evaluation event, not another tuning opportunity. The material does not certify production readiness, accessibility, learner outcomes or a professional qualification.