The FourierFlow is a simple, yet powerful tool for generating synthetic time series data using deep generative models. The model is slightly modified and applied to the Market Scenario Generator Hackathon: From Stability to Storms (for more details of the hackathon, visit Hackathon Website).
Clone this repository to your local machine. I recommend to create a new environment by running
conda env create -f environment.yml- Navigate to the config_ff.yaml file.
- Adjust the parameters according to your needs:
input_dim: Dimensionality of input features.output_dim: Dimensionality of output features.hidden_dim: Size of hidden layers.n_flows: Number of flow layers.n_lags: Number of time lags.vol_activation: Activation function for volatility modeling (e.g., “softplus”).- Other hyperparameters (batch size, learning rate, etc.).
- Pretraining or Checkpoint:
- Choose between two modes:
- Pretrain:
- Load your training data and labels (regular and crisis data).
- Train the generator using generator_regular.fit() and generator_crisis.fit().
- Save the combined model dictionary using save_combined_model_dict().
- Checkpoint (Pickle Files):
- Load the pre-trained model from model_dict.pkl.
- Pretrain:
- Choose between two modes:
- Generating Samples:
- Run the script main.py in console:
python main.py - Specify the condition (e.g., crisis or regular) by setting
condition[0]in themain.py. - The generated synthetic data will be saved to a pickle file.
- Run the script main.py in console:
[1] Alaa, A.M., Chan, A.J., & Schaar, M.V. (2021). Generative Time-series Modeling with Fourier Flows. International Conference on Learning Representations.