.
├── CFG-Syn/             # Syntax-validity processing engine
├── CFG-Sem/             # + Semantic-validity processing engine
├── CFG-Sem-k/           # + Constraint-optimized processing engine
├── RPN/                 # Reverse Polish Notation calculator
├── ML/                  # Baseline methods that directly use machine learning techniques to predict stock returns
├── requirements.txt   # Project dependencies
└── README.md          # Project documentation

The framework is shown below.

• Python 3.8–3.11 (Python 3.9 or 3.10 recommended)

• pip (Python package manager)

1. Clone the repository:

   git clone <your-repo-url>
   cd alpha-discovery-via-grammar-guided-learning-and-search

2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # Linux/macOS
   # or
   .\venv\Scripts\activate   # Windows

3. Install dependencies:

   pip install -r requirements.txt

4. Verify the installation:

   python -c "import torch; print(f'PyTorch version: {torch.__version__}'); print(f'CUDA available: {torch.cuda.is_available()}')"

You can obtain the required stock data using one of the following approaches.

Run the built-in script of AlphaGen to automatically fetch and preprocess data using Baostock:

python alphagen/data_collection/fetch_baostock_data.py

Alternatively, you may leverage qlib for more flexible and large-scale financial data handling.

To execute the integrated module pool, just enter the folder directory of different projects. For example:

cd CFG-Sem-k

and run:

python run_pool.py

This project includes several representative baseline methods for comparison.

• AlphaGen: AlphaGen

• AlphaQCM: AlphaQCM

• ML: Pure machine learning baselines that directly predict stock returns without explicit symbolic alpha construction.
  To run these baselines, execute:

  cd ML
  python main.py

The result of backtest is shown below.

• Installation may take some time — ensure a stable network connection.

• The project depends on large libraries such as PyTorch and TensorFlow — ensure sufficient disk space.

• For CUDA-related issues, please verify GPU driver and CUDA version compatibility.

If you encounter any issues, please submit an issue on GitHub or contact the development team.