Dictionary of Applied Machine Learning

Dictionary of Applied Machine Learning — companion repository

Companion code for the Dictionary of Applied Machine Learning (course edition, CC BY 4.0): 107 machine-learning terms, each defined precisely and cross-referenced, with citations, a typeset PDF, and a Python demo that recomputes what the entry states.

Directory	Contents
notebooks/	The per-term Python demos as Jupyter notebooks — each opens in Colab with one click from its demo page
mldict/	mldict.sty: the dictionary's notation as a LaTeX package (\usepackage{mldict})
mcp/	An MCP server exposing the dictionary to AI assistants (Claude Code, VS Code Copilot, …)

The notebooks and mldict.sty are generated from the dictionary's source; edits belong there, not here. The dictionary itself — definitions, symbol list, per-term PDFs — lives at dictionaryofml.org; the full content is also machine-readable as terms.json and indexed for LLM crawlers in llms.txt.

Provenance and disclaimer

The source code in this repository (demo notebooks, LaTeX package, MCP server) was created with the help of large language models (LLMs), under review by the author. It is provided as is, without warranty of any kind: use at your own risk and verify results independently before relying on them.

Basic checks are applied to the Python demos before release:

• each demo runs its own check() assertions on the quantities it computes, with fixed random seeds for reproducibility;
• ruff check --select F (correctness rules only: undefined and unused names);
• bandit, restricted to the exec/eval/shell-injection checks (B102, B307, B602, B605) — the one deliberate exception is the llm demo, whose subject is an LLM wrapper executing emitted code;
• project-specific linters for figure style and for keeping a demo's prose within the concepts its dictionary entry introduces.

These are screens, not proofs of correctness.

The demos import two third-party packages, numpy and matplotlib (both preinstalled on Colab); the source repository pins their exact versions, since the committed figure data is byte-reproducible only under the pinned pair.

MCP server

pip install -r mcp/requirements.txt
claude mcp add --scope user dictionaryofml -- python /path/to/mcp/server.py

Tools: list_all_terms, lookup_term, search_terms, get_related_terms. Content is fetched from the site's terms.json and cached for a day.

Citing

@misc{dictml,
  author = {Jung, Alexander and Olioumtsevits, Konstantina and Schnoor, Ekkehard},
  title = {Dictionary of Applied Machine Learning (course edition)},
  doi = {10.5281/zenodo.21569296},
  note = {ISBN 978-952-64-3013-3, CC BY 4.0},
  url = {https://dictionaryofml.org}
}

Each term page carries its own "Cite this entry" BibTeX block. This repository is a course-edition companion, distinct from the copyrighted Springer Dictionary of Applied Machine Learning.

License

CC BY 4.0. Maintained by Alexander Jung, Aalto University; corrections and suggestions are welcome at the address on dictionaryofml.org.