"Learning Deep Representations of Data Distributions"

This repository is the source code for the book "Learning Deep Representations of Data Distributions". If you just want to read the book, you should not need this repository. We have a copy at this link that will be updated periodically.

Generally, you should be accessing the source code for one of the following purposes:

• You want to build the book, or one of its chapters, from scratch. See this section.
• You want to use the source code that generated a figure in the book. See this section.
• You want to edit the website. See this section.
• You want to ask a question, or tell us that something in the repository doesn't work quite right. See this section.
• You want to contribute some content, for example a translation, or some technical content within one or more chapters. See this section.

Building the Book or a Chapter

Prerequisites for Building the Book or a Chapter

• A local pdflatex distribution (TexLive is a recommended choice)
• A binary of latexmk, which comes with TexLive but can alternatively be installed using brew

How to Build the Book or a Chapter

To build the book, navigate to the repository base folder and run:

latexmk book-main.tex

and read book-main.pdf.

You can also build individual chapters via latexmk! To build a chapter, say Chapter 2, you can navigate to chapters/chapter2 and run:

latexmk classic-models.tex

and read classic-models.pdf. (classic-models.tex is the chapter file for Chapter 2). We did some LaTeX hacks so that compiling the main book or individual chapters should work in your IDE (please file an issue if it doesn't work).

You can do the same with the Chinese translation of the book (though note you may need to use latexmk book-main_zh.tex -f, etc., the force flag being due to missing characters).

Using the Code

Prerequisites for Using the Code

• The Python package manager uv

How to Use the Code

• Navigate to the repository and run uv sync, which creates a Python virtual environment restricted to the repository.
• Go to the chapter folder that you are interested in and navigate to the code directory, i.e., chapters/chapter3/code.
• Run the desired file with uv run <filename> and observe the outputs.

Building the Website

• A local pdflatex distribution (TexLive is a recommended choice)
• The Python package manager uv
• A latexml installation which can be obtained in several ways
• An installation of ImageMagick

How to Build the Website

• Navigate to the repository and run uv sync, which creates a Python virtual environment restricted to the repository.
• Go to the website/latex_to_html directory and run uv run latex_to_html_converter.py ../../book-main.tex ../html. This should take a long time (around 10-30 minutes).
• Navigate one level up and into the newly created html directory.
• Start an HTTP server using uv run python3 -m http.server <PORT>. (The constant <PORT> represents a free port and is otherwise completely arbitrary; one possibility is 13579.)
• Navigate to http://localhost:<PORT>/index.html.

Note: The AI helper calls a variety of different models by making queries to a Cloudflare Worker proxy. We use this proxy because we do not want to expose API keys to everyone. As such, the worker will unfortunately not be open-sourced. If you really need access, talk to Druv.

Note: To build the website using latex_to_html_converter.py with a working bibliography, you need to make sure the book-main.bbl file is up to date, which may mean that you have to build the book first.

Raising an Issue

Prerequisites for Raising an Issue

A GitHub account.

How to Raise an Issue

Use the GitHub "Raise Issue" UI, linked here. Please give us as much detail as possible when making your issue.

• If you are raising an issue about any of the code, please let us know your system details: OS, Python version, uv version, the configuration of any GPUs/TPUs you are using, etc.
• If you are raising an issue about the content, please be as descriptive as possible. If you believe that something is wrong, please give a concrete reason why (e.g. a counterexample).

Please use English as this maximizes the number of people who can help.

Making a Contribution

Prerequisites for Making a Contribution

• Minimally, a text editor and GitHub account.
• But you probably also want to be able to build the book and/or chapter and/or edit and run the code.

How to Make a Contribution

• Make a new branch on GitHub, and call it whatever you like (among names that are not taken), e.g., my_new_branch.
• Clone the repository locally using git clone:

git clone https://github.com/Ma-Lab-Berkeley/deep-representation-learning-book.git

• Change the branch, i.e.,

git checkout -b my_new_branch

• Make whatever edits you want, then add them, commit, and push to GitHub:

git add <some files you changed>
git commit -m "<descriptive commit message>"
git push -u origin my_new_branch

• You can commit and push as many times as you want until you are satisfied.
• Make a pull request to main using the GitHub UI. The pull request will likely go through some revision from a core contributor, and may be merged to main afterwards.

Please use English as this maximizes the number of people who can attend to your contribution.

General Guidelines for Contributing

• For code: please examine current code files and see the coding patterns.
• For LaTeX: please examine current LaTeX files and see the coding patterns.

Citation Information

Many thanks!

@book{ldrdd2025,
  title={Learning Deep Representations of Data Distributions},
  author={Buchanan, Sam and Pai, Druv and Wang, Peng and Ma, Yi},
  month=aug,
  year={2025},
  publisher={Online},
  note={\url{https://ma-lab-berkeley.github.io/deep-representation-learning-book/}.}
}