ML-2021-notes: A Student's Written Companion to Hung-yi Lee's 2021 Machine Learning Course
Lecture notes for Professor Hung-yi Lee's Machine Learning (2021 Spring) course at National Taiwan University (NTU)
At a glance
- What is it?
- The repository is a set of Chinese-language lecture notes for NTU's Machine Learning 2021 Spring course, published in three formats. It is a study aid tied to a specific video series, not a library, and the README itself flags that parts of it are incomplete.
- Who is it for?
- Adopt it if you are working through the Hung-yi Lee 2021 Spring playlist and want a written Chinese-language companion that maps chapter by chapter onto the videos, since the README links each note to its source lecture. Do not adopt it as a standalone English-language textbook or as a source of runnable code, because the README describes notes and slides, not exercises or implementations.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository last received commits 55 days ago.
- What is it written in?
- GitHub does not report a main language for this repository.
Answers come from the project's GitHub data, last synced on October 9, 2026, and from our analysis. They are not legal advice.
Editorial analysis
What the Notes Cover and Who They Are Written For
The repository is a set of lecture notes for the course 機器學習 (Machine Learning) 2021 Spring, taught by 李宏毅 (Hung-yi Lee) at 臺灣大學 (NTU). The README states plainly that the notes were made by digesting and understanding the instructor's class content, and that the author welcomes corrections where the understanding is wrong. That sentence sets the expectation: this is one student's interpretation of a lecture series, not an official course artifact.
The audience is narrow and specific. If you are following the linked YouTube playlist and read Chinese, the notes give you a written thread to follow alongside the videos. The README lists chapters 01 through 08 and beyond, covering basic machine learning concepts, deep learning guidance and optimization, CNNs, self-attention, transformers, GANs, self-supervised learning with BERT, and auto-encoders. The topic tags on the repository match that list.
What it is not: it is not a framework, not a set of notebooks, and not a replacement for the lectures. Nothing in the README describes runnable code, exercises, or datasets. Anyone arriving from a search for a machine learning library will find a document collection instead.
Three Publication Formats and How They Diverge
The README offers the same material through three channels. Notion pages host the primary online version. A GitHub Pages site at chsiang426.github.io/ML-2021-notes serves an HTML rendering. A PDF directory in the repository holds downloadable files. Each chapter row in the README table carries links to all three where they exist.
That redundancy has a practical consequence for anyone citing or archiving the notes. The three copies can drift. The README does not state a generation pipeline, so there is no documented guarantee that the Notion page, the HTML file, and the PDF for a given chapter contain identical text. The HTML links use URL-encoded Chinese characters, for example the chapter 03 path contains 卷積神經網路 in percent-encoded form, which means the file names themselves are language-dependent. If you script a download against those paths, expect to handle encoding.
The PDF links point into the repository's PDF folder, so those files travel with the git history. The Notion and GitHub Pages versions live outside it. Cloning the repository gives you the PDFs and whatever source the site is built from, but not the Notion content.
Chapter Mapping to the Video Playlist
The README table is organized by chapter number, topic, note link, and video link. Chapter 02 is split into three sub-notes: General Guidance, 類神經網路優化技巧 (neural network optimization techniques), and Loss of Classification. Each sub-note has its own Notion, Website, and PDF links, and the video column lists six separate YouTube links for that chapter. Chapter 01 maps to two videos. Chapter 03, on CNNs, maps to one.
That mapping is the most useful structural feature of the repository. It lets you move from a lecture timestamp to a written summary without searching. It also reveals the granularity mismatch: some chapters are one video and one note, others are six videos compressed into three notes. The README does not explain how the author decided where to split.
The video links use youtu.be short URLs, so they are stable enough to bookmark but carry no timestamp parameters. You still have to find the relevant moment in the lecture yourself.
The Incompleteness the README Admits
The third reminder in the README states that several small parts of the notes are incomplete, where the instructor covered material the author did not write down. This is the single most important caveat in the repository, and it is stated without qualification.
Because the notes are a personal digest, gaps are structural rather than accidental. A missing derivation or a skipped example is not a bug to be filed; it is a boundary of the source. If you are using the notes to prepare for an exam or to fill a knowledge gap, you cannot assume that silence on a topic means the topic was not taught. The README's own framing, that the content follows the class and that errors should be pointed out, puts the burden of verification on the reader.
The absence of any stated review process compounds this. There is no mention of a second reader, an errata file, or a changelog. Corrections appear to arrive through whatever channel the reader chooses, and the README does not name one.
Images, Attribution and the Missing Licence
The README says the images in the notes are captured from the instructor's class slides, which come from the instructor's course website at speech.ee.ntu.edu.tw. The repository metadata does not state a licence. Neither the README excerpt nor the repository description names one.
That combination matters if you plan to redistribute the notes, embed them in internal training material, or republish them on another site. The text appears to be the author's own writing, but the images are attributed to a third party's slides. Without a stated licence on the repository, the default position is that no permission has been granted for reuse beyond whatever the hosting platforms allow. The README does not address this, and I cannot resolve it from the available material.
For personal study, the question is academic. For anything else, treat the image provenance as an open item and check the instructor's course site for its own terms before copying anything.
How This Differs From a Textbook or a Code Repository
The obvious alternative for someone who wants structured machine learning material is a textbook, for example a standard university text that comes with exercises and a defined edition history. The difference in approach is not quality but contract. A textbook promises coverage and a stable artifact; these notes promise a companion to a specific set of lectures and explicitly disclaim completeness.
A second alternative is the lecture series itself. The README links every chapter to its source videos, which means the notes are downstream of the videos by design. If you can follow the lectures directly, the notes add compression and a written index, not new information. If you cannot follow the lectures, for example because of language or time, the notes are a lossy summary and the gaps the README admits will be the parts you most need.
A third alternative is a code-first course repository with notebooks you can execute. Nothing in this repository's material suggests it plays that role. There is no setup section, no dependency list, and no mention of running anything.
Maintenance and What to Check Before Depending on It
The repository is not archived, and the last push date in the metadata is 2026-08-16. No releases are listed. The README contains no versioning scheme, no changelog, and no stated update cadence. The notes correspond to a 2021 Spring course, so the underlying lectures are fixed; the repository's own activity is the only variable.
That is a low-maintenance shape. There is no build to break and no dependency to upgrade. The costs that remain are the ones already described: the licence is unstated, the images come from a third party's slides, and the README admits incomplete sections. Anyone deciding whether to depend on the notes should start by opening one chapter in all three formats and comparing them, then checking the corresponding video for a section the README flags as thin. If the three copies agree and the gaps are acceptable for your use, the repository does what it says. If they diverge, the PDF folder is the version that travels with the git history, which makes it the one to pin against.
Editorial conclusion
Adopt it if you are working through the Hung-yi Lee 2021 Spring playlist and want a written Chinese-language companion that maps chapter by chapter onto the videos, since the README links each note to its source lecture. Do not adopt it as a standalone English-language textbook or as a source of runnable code, because the README describes notes and slides, not exercises or implementations. Verify two things before relying on it: the licence, which the repository metadata does not state, and the slide image permissions, since the README says the images are taken from the instructor's own course slides. Then check the incomplete sections against the videos yourself.
Frequently asked questions
Can you provide some notes for learning machine learning?
This repository holds notes for Professor Li Hongyi's 2021 spring Machine Learning course at National Taiwan University, published as Notion pages, pages on the project website, and downloadable PDFs, with a table that links every chapter to its lecture videos.
Which chapters does ML-2021-notes cover?
The table runs from 01 on basic machine learning concepts, through a DeepLearning chapter split into general guidance, neural network optimisation techniques and loss of classification, then CNN, self-attention, Transformer, generative adversarial networks, self-supervised learning with BERT, and auto-encoders.
Where can I read the ML-2021 notes?
Each chapter row carries three links to the same material: a Notion page, a page on the project website at chsiang426.github.io, and a PDF that can be downloaded from the PDF directory of the repository.
Are the ML-2021 notes complete?
The author says a few small parts are incomplete, where the professor covered material in the lecture but it was not written down. The README also invites corrections where the notes misunderstand what was taught.
Where do the images in the ML-2021 notes come from?
All images are captured from the professor's lecture slides, and those slides come from the course website maintained by the lecturer at speech.ee.ntu.edu.tw.
Is ML-2021-notes still being worked on?
The repository publishes no GitHub releases and holds only .gitignore, PDF/, README.md and docs/ at the top level, so there is no software to build. Its last push was on August 16, 2026, well after the 2021 spring term it documents.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/chsiang426-ml-2021-notes)