COMP 4222 Group 22 Project

1. Project File Structure

• data/

  • raw/
    • README.md
    • HIV_train_oversampled.csv
    • test.csv

• GNN/

  • data_preprocessing.py
  • hyper_parameters.py
  • main.py
  • model.py
  • test.py
  • train.py
  • utils.py

• CNN/

  • featurizer.py
  • graph_sage_embedding.py
  • model.py
  • train.py

• SVM/

  • train.py

  2. General Purpose of Each File

• data/

  • raw/ Contains raw molecular data stored in CSV format.
    • HIV_train_oversampled.csv: Dataset with oversampled instances to address class imbalance.
    • test.csv: Dataset for testing and validation purposes.

• GNN/

  • data_preprocessing.py: Implements a graph converter that transforms molecular strings into graph-structured data using PyTorch Geometric, storing the processed data in data/processed.
  • hyper_parameters.py: Stores hyperparameters for the GNN model for easy tuning.
  • main.py: Main script for training and evaluating the GNN model.
  • model.py: Defines the architecture and implementation of the GNN model.
  • test.py: Provides testing functionality for evaluating the model performance.
  • train.py: Implements training functionality for the GNN model.
  • utils.py: Contains utility functions for evaluating the model's performance, such as calculating confusion metrics and determining the total number of model parameters.

• CNN/

  • featurizer.py: Provides functions for converting graph objects into embeddings using GraphSAGE.
  • graph_sage_embedding.py: Defines the architecture of GraphSAGE.
  • model.py: Defines the architecture of a 1-dimensional CNN.
  • train.py: Contains the main logic for training the CNN model.

• SVM/

  • train.py: Contains the main logic for fitting the SVM model.

• requirements.txt: Specifies the packages and their versions used in this project.

3. Usage

Setting Up the Environment

1. Optional: Setting up a virtual environment:

   python -m venv .venv
   # Activate the virtual environment
   .venv\Scripts\activate

2. Installing required packages:

   pip install -r requirements.txt

Training GNN

1. Navigate to the GNN directory:

   cd GNN

2. Run the main script for training the GNN model:

   python main.py

   Note: For the first run, it may take some time to convert the molecular string into a graph.

Training CNN

1. Navigate to the CNN directory:

   cd CNN

2. Run the training script for the CNN model:

   python train.py

   Note: Encoding the graph using GraphSAGE before training the CNN model may take some time.

Fitting SVM

1. Navigate to the SVM directory:

   cd SVM

2. Run the ftiiting script for the SVM model:

   python train.py

   Note: Encoding the graph using GraphSAGE before fitting the SVM model may take some time.

---