run_graph_experiments.sh

#!/bin/bash

# Run graph classification experiments with hyperparameters from Appendix C.1 (Table 3)
# Use the hgt conda environment
PYTHON="\path\to\your\conda\envs\python.exe"

# Mutagenicity - M=200, R=6, Θ=1.4, Lr=1.1e-2, Wd=5e-4
echo "Training on Mutagenicity (10-fold CV)..."
$PYTHON scripts/train_graph.py \
  --dataset mutagenicity \
  --cuda 0 \
  --n_folds 10 \
  --lr 1.1e-2 \
  --weight_decay 5e-4 \
  --M 200 \
  --R 6 \
  --theta 1.4 \
  --batch_size 32 \
  --epochs 1000 \
  --save_model

# Proteins - M=50, R=5, Θ=1.9, Lr=3.2e-3, Wd=5e-3
# echo "Training on Proteins (10-fold CV)..."
# $PYTHON scripts/train_graph.py \
#   --dataset proteins \
#   --cuda 0 \
#   --n_folds 10 \
#   --lr 3.2e-3 \
#   --weight_decay 5e-3 \
#   --M 50 \
#   --R 5 \
#   --theta 1.9 \
#   --batch_size 32 \
#   --epochs 1000 \
#   --save_model

# Mutag - M=200, R=7, Θ=1.2, Lr=3.7e-2, Wd=5e-3
# echo "Training on Mutag (10-fold CV)..."
# $PYTHON scripts/train_graph.py \
#   --dataset mutag \
#   --cuda 0 \
#   --n_folds 10 \
#   --lr 3.7e-2 \
#   --weight_decay 5e-3 \
#   --M 200 \
#   --R 7 \
#   --theta 1.2 \
#   --batch_size 32 \
#   --epochs 1000 \
#   --save_model

# NCI1 - M=50, R=7, Θ=3.7, Lr=1.6e-2, Wd=5e-4
# echo "Training on NCI1 (10-fold CV)..."
# $PYTHON scripts/train_graph.py \
#   --dataset nci1 \
#   --cuda 0 \
#   --n_folds 10 \
#   --lr 1.6e-2 \
#   --weight_decay 5e-4 \
#   --M 50 \
#   --R 7 \
#   --theta 3.7 \
#   --batch_size 32 \
#   --epochs 1000 \
#   --save_model

# PTC_MR - M=50, R=5, Θ=1.7, Lr=2.9e-3, Wd=2.5e-5
# echo "Training on PTC_MR (10-fold CV)..."
# $PYTHON scripts/train_graph.py \
#   --dataset ptc_mr \
#   --cuda 0 \
#   --n_folds 10 \
#   --lr 2.9e-3 \
#   --weight_decay 2.5e-5 \
#   --M 50 \
#   --R 5 \
#   --theta 1.7 \
#   --batch_size 32 \
#   --epochs 1000 \
#   --save_model

echo "Experiment completed!"