ml-road: the curated course table that maps a machine learning education

The courses table as the core artifact

The repository's heart is a table with five columns, course name, institution, lecturer, links and category, and the rows read as a who's who of machine learning pedagogy. Andrew Ng appears three times, the Coursera Machine Learning course with Bilibili and YouTube links, Stanford's original Machine Learning open course with a Netease mirror, and the deeplearning.ai Deep Learning specialization. National Taiwan University's Hsuan-Tien Lin contributes both Machine Learning Foundations and Machine Learning Techniques, each with Bilibili and YouTube recordings. The links' geography is telling, Bilibili and Netease mirrors beside the YouTube originals, a bilingual accessibility choice that reflects the audience the collection was assembled for. The lecturer column is the table's quiet quality signal, courses are listed under the people who taught them, Ng, Lin, Li, Manning, Levine, Huyen, Jurafsky, so a learner follows instructors across institutions rather than brands, and the repeated names across rows reveal which teachers the collection's author judged worth following more than once.

The Stanford and Berkeley specializations

The deep learning corner of the table is anchored by the Stanford and Berkeley courses every practitioner recognizes. CS231n, Convolutional Neural Networks for Visual Recognition, taught by Fei-Fei Li, with homepage and YouTube playlist links and categorized as deep learning and computer vision. CS224n, Natural Language Processing with Deep Learning, taught by Christopher Manning. Berkeley's CS 294, Deep Reinforcement Learning, taught by Sergey Levine. And Stanford's CS 20, TensorFlow for Deep Learning Research, taught by Chip Huyen, notable for linking its own GitHub repository of tutorials alongside the homepage. The category column groups these, and the groupings make an emergent curriculum visible, vision through CS231n, language through CS224n and CMU's Neural Networks for NLP by Graham Neubig, decision making through CS294. The YouTube playlists linked beside the homepages are also a deliberate completeness choice, official course sites move or rewrite each year, while the recorded lecture playlists remain stable, so the table survives its sources by pointing at both.

Beyond the famous four

The table's depth is in its less famous rows. Oxford University's Deep Learning for Natural Language Processing by Phil Blunsom, with a slides repository on GitHub beside the homepage. NTU's Applied Deep Learning, jointly taught by Yun-Nung Chen and Hung-Yi Lee. NYU's Mathematics of Deep Learning by Joan Bruna, linking the course's GitHub. Dan Jurafsky and Chris Manning's Introduction to NLP at Stanford. ChengXiang Zhai's Text Mining and Analytics from UIUC on Coursera. And Google's Machine Learning Crash Course with TensorFlow APIs. Mathematics appearing as its own entry is the table's quiet argument, the gap between following tutorials and understanding the field is a math gap, and Bruna's course is the listed remedy. The Google crash course entry rounds out the list with an option for the impatient, a shorter, official, hands-on path that fits between the semester-length university courses, and its TensorFlow API focus dates it as an entry point rather than an endpoint, the role the table assigns it.

The disclaimer that shapes the repository

The disclaimer section is direct about the collection's legal position, the resources in this repository are only for educational purpose, do not use them for any form of commercial purpose, and a second clause addresses the edge the repository walks, if the author of an ebook finds their intellectual property violated because of contents in this repository, they should make contact and the relevant material will be removed as soon as possible. Resource collections that link hosted ebooks live under this exact pressure, and the repository's answer is a standing removal policy rather than denial. The MIT license on the repository's own code sits beside this, the two documents together defining what is free to reuse and what only points elsewhere.

Practice alongside the pointers

The repository structure pairs resources with practice, a projects directory and a resources directory beside the readme, and the primary language registered on GitHub is Python, the language of the practice code rather than the markdown tables. The combination is the repository's stated scope, machine learning resources, practice and research, the three nouns of its description. A learner following the courses table lands in the projects directory for the implementation half, and the pairing matters because the most common failure of self-directed ML education is consuming courses without building, the structure nudging against that failure mode. For a self-learner the division of labor is worth following literally, watch the lecture from the resources side, implement in the projects side, and the Python language registration on the repository confirms where the runnable half lives.

From machine learning to agentic AI

The repository's description now reads machine learning and agentic AI resources, practice and research, the agentic addition marking the collection's expansion beyond classical ML into the current agent era. The last push landed 2026-08-01, and there are no GitHub releases, the collection shipping by commit as resource lists do. The star history chart at the top of the readme shows the repository's long visibility in the community, and its continued presence on curated lists is the compounding effect a well-maintained index earns, each new cohort of learners finding it while searching for exactly the sequencing the table provides.

Editorial conclusion

Use ml-road as a starting map when structuring a machine learning education, since its course table sequences the canonical classes by institution and topic and its project directories hold practice code to pair with the theory. Treat it as an index rather than a curriculum, no schedule or assessments live here, and honor the disclaimer, resources are for educational purposes only with commercial use prohibited. Verify links as you go, the table mixes Bilibili, YouTube, Netease and course homepages that age at different rates, and read the table's category column to assemble a path across machine learning, deep learning, NLP and reinforcement learning rather than consuming it alphabetically.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. README
4. yanshengjia/ml-road on GitHub