This is the official repository for the 2025 Bootcamp on Interpretable and Explainable AI by Vector Institute. All reference implementations are present in implementations folder. Small datasets are present in the corresponding subfolders, and the large datasets are present on the cluster at /ssd003/projects/aieng/public/interp_bootcamp/datasets.

• implementations folder contains all the code covered in this bootcamp, and is organized by topic. Each topic has its own directory containing python scripts, notebooks, datasets and a README for guidance.
• scripts contains utility scripts for submiting training jobs.
• pyproject.toml file configures various build system requirements and dependencies, centralizing project settings in a standardized format.
• requirements.txt file contains a list of packages and their versions required to run the implementations. This can be used to setup your own environment in the cluster/your machine.
• uv.lock file contains list of project dependencies generated by uv.

Each topic within the bootcamp has a dedicated directory in the implementations/ folder. In each directory, there is a README.md file that provides an overview of the topic, prerequisites, and notebook descriptions.

Here is the list of topics covered in this bootcamp:

• Post-hoc Explainability Methods in Post-hoc:

  • LIME
  • SHAP
  • PDP
  • ALE
  • Integrated Gradients
  • Counterfactual Explanations

• Interpretable models in Interpretable-models:

  • GAM
  • NAM
  • EBM
  • GOSDT/GHOUL
  • COXNAM
  • NBM
  • ProtoPNet
  • TabNet
  • Self Attention
  • B-COS Networks

Please check the following requirements before running the code examples on your cluster.

1. A Github account with security credentials setup to clone the repository.
2. Access to the Vector Institute's cluster: You will be able to login via ssh username@v.vectorinstitute.ai and 2FA via mobile authentication.
3. VPN setup via Fortinet client: After connecting via VPN, you will be able to login to the Jupyter Hub server hosted at https://vdm1.cluster.local:8000/.

Check this if you are running the code on your laptop.

• Python >= 3.10

1. As a part of your cluster account, you have been allocated a scratch folder where code, checkpoints and training artifacts can be stored. It is at the location

/scratch/ssd004/scratch/<username>

It can also be a different path with ssd003.

2. A virtual environment is present at a shard location on the cluster. Please source it using the following command:

source /ssd003/projects/aieng/public/interp_bootcamp/venv/bin/activate

3. Large datasets used within the repository are present at the location /ssd003/projects/aieng/public/interp_bootcamp/datasets.
4. Pretrained models are available at the location /ssd003/projects/aieng/public/interp_bootcamp/checkpoints.

Please follow the steps in instructions.md to run the examples in this repository. Please follow steps in scripts.md to run slurm scripts.

This project is licensed under the terms of the LICENSE.md file located in the root directory of this repository.

To get started with contributing to our project, please read our CONTRIBUTING.md guide.

For more information or help with navigating this repository, please contact Dhanesh Ramachandram, Applied ML Scientist, Health Lead at [dhanesh.ramachandram@vectorinstitute.ai] or Ananya Raval, Software Developer, AI Eng. [ananya.raval@vectorinstitute.ai]