This is the final project repository template for Machine Learning with Probabilistic Programming.
Please follow these instructions to make a copy of this repository and push it to your own GitHub account.
Make sure to create a new repository on your own GitHub account before starting this process.
We have included a example of a Jupyter notebook under
/notebook-example/example.ipynb. This shows how to use markdown along with
LaTeX to create section headings and typeset math.
Your final project notebook should go under
/final-project/final-notebook.ipynb. This notebook will be your final report.
We must be able to run it in a reasonable amount of time. (If your project
involves a massive dataset, please contact me.)
Your final report should be 8 pages long. Since it is hard to translate between a Jupyter notebook and page numbers, we've come up with the following metric:
the Markdown export of your notebook should be approximately 1500 words.
To compute this, save your Jupyter notebook as a Markdown file by going to
File > Download as > Markdown (.md)
and then counting the words
wc -w final-notebook.md
Since this includes your code as well, we encourage you to develop separate python scripts and include these in your final notebook. My recommendation is that you only do basic data loading, manipulation, and plotting within Jupyter; do all of the heavy lifting in separate Python files. (Note our strict guidelines on coding style below.)
Your notebook should follow the basic structure described in the project proposal template. Make sure to clearly indicate section headings and to present a clear narrative structure. Every subsection of your report should correspond to a particular step of Box's loop. Feel free to include images; you can embed them in markdown cells.
Use Python 3.9.X.
Configure a virtual environment. Follow the documentation here.
Once you activate the virtual environment, use pip to install a foundational set of
packages.
(venv)$ pip install -r requirements.txt
This should install CmdStanPy, along with Jupyter and other useful libraries.
Please install cmdstan into its default location ~/.cmdstan by following the instructions here: Installing CmdStan
If you introduce any new dependencies to your final project, you MUST
update requirements.txt with pinned versioning.
There is a comprehensive .gitignore file in this repository. This should prevent you from committing any unnecessary files. Please edit it as needed and do not commit any large files to the repository. (Especially huge datasets.)
Any additional code you write must pass flake8 linting. See this
blog post for details.
The first thing we will do after cloning your repository is:
(venv)$ flake8
If your repository fails any checks, we will deduct 20% from your final project grade.