eroell/ellis-summerschool

Graphical Models Tutorial, Using GitHub Codespaces

Setup

1. Fork this repository
2. Click on the green 'code' button
3. Navigate to 'Codespaces'
4. Click the '+' button to create a codespace based on your forked repository
5. Run all cells in notebooks/setup.ipynb

The last step will add the relevant packages and data to your codespace. The actual tutorial is in the file notebooks/experiment.ipynb, but do not start going through that until told.

Blurb

In this repository, we will investigate the following dataset:

• Spatio-temporal immune zonation of the human kidney, with an associated paper.

We will investigate it with two graphical models:

1. GLasso
   • Makes independence assumption on the samples
   • Uses regularization for more robust inference
2. GmGM
   • Does not make an independence assumption; finds graphs for both genes and cells
   • Uses thresholding rather than regularization