[P] Open-source ML homeworks with auto-tests - fundamental algorithms from first principles
A professor publishes a complete set of hands-on ML assignments with automated testing to build deep understanding.
A professor from Skoltech (Russia's premier science and technology institute) has publicly released the complete homework assignments from a machine learning course, designed to address a gap he identified in his own education: a lack of deep, hands-on understanding of core algorithms. The philosophy is that true comprehension comes from building replicas from scratch. To make this practical, the course provides structured Jupyter notebooks that guide students through implementing fundamental ML and deep learning concepts step-by-step, preventing the 'terror of a blank page' while ensuring foundational learning.
Crucially, the course borrows from industry software development practices by implementing comprehensive automated testing. Students receive a starter template and a test suite, and must write code that passes these tests to complete the assignments. This system provides immediate feedback, automates grading to reduce teaching staff burden, and teaches students to debug based on error messages. The full package, available on GitHub, includes notebooks, helper scripts, auto-tests, grading scripts, and pre-generated test data, all under a permissive license for broad reuse in other educational contexts.
- Course materials include Jupyter notebooks and auto-tests for building ML algorithms from first principles.
- Uses automated testing for immediate student feedback and to reduce grading workload on teaching staff.
- Full set of assignments is open-source on GitHub under a permissive license for educators and learners.
Why It Matters
Provides a scalable, hands-on curriculum for teaching foundational ML, valuable for educators and self-learners worldwide.