How maneeshks/homemade-machine-learning works
Project type and purpose
• This is an open-source educational library of homemade machine learning algorithms with interactive Jupyter Notebook demos.
• It is not a user-facing “application” but rather a code library + demo suite intended to teach and illustrate core ML algorithms in Python.
File structure
• Top-level directories:
• data/ – CSV datasets used by demos
• homemade/ – Python package implementing algorithms (subpackages: linear_regression, logistic_regression, k_means, anomaly_detection, neural_network, utils)
• notebooks/ – Jupyter Notebook demos that import from homemade/ and data/
• Configuration: requirements.txt, pylintrc, .travis.yml, GitHub metadata (CONTRIBUTING, CODE_OF_CONDUCT, LICENSE)
README for insights
• Purpose: show math behind each algorithm by implementing from scratch
• Integration points: Jupyter NBViewer for static demos, Binder for runnable environments, Travis CI for continuous integration
Detail is critical
• Utils/features and utils/hypothesis submodules as inputs to algorithm modules
• Explicitly show notebooks grouped by algorithm area pointing back at the modules they import
• Config files and CI pipeline as part of the maintenance/quality assurance component