Graph Deep Learning Project – TopoMasking

This repository contains the implementation and report for the Graph Deep Learning group project on Linear Transformer Topological Masking (TopoMasking), based on the ICLR 2025 paper “Linear Transformer Topological Masking with Graph Random Features”.

---

📦 1. Clone the Repository

To get started, clone this repository locally:

git clone https://github.com/USI-Projects-Collection/GDL_Project.git
cd GDL_Project

---

⚙️ 2. Conda Environment Setup

We use Conda to manage dependencies and ensure reproducibility

▶️ Create the environment

conda env create -f environment.yaml
conda activate topo_masking

If the environment file changes (e.g., someone added new dependencies) see Section 5:

conda env update -f env/environment.yaml --prune

---

📄 3. LaTeX Report Compilation

Use a Makefile to build the final PDF.

🔹 On macOS

brew install --cask mactex

🔹 On Ubuntu / Debian

sudo apt install texlive-full

---

🧱 Build the PDF

From inside the Template_GDL_Report/ folder:

make          # Compiles the LaTeX report
make clean    # Removes temporary LaTeX files

---

👥 4. Collaboration Workflow

When adding new dependencies:

1. Install the package in your Conda environment:

   conda install nome_pacchetto

2. Add the package manually to the dependencies section of environment.yaml:

   dependencies:
     - pytorch
     - pyg
     - numpy
     - package_you_installed # newly added package

3. Commit and push the updated file so everyone can sync with:

   conda env update -f environment.yaml --prune

---

🧾 5. License & Credits

© 2025 — Università della Svizzera italiana (USI), Master in Artificial Intelligence.
Developed by the Graph Deep Learning project group.