gsnn.gsnn.simulate.datasets

Functions

simulate(G, n_train, n_test, input_nodes, ...)

Generate samples from a synthetic graph-structured data-generation process.

simulate_10_in_25_func_10_out_cyclic(...[, ...])

Create a complex cyclic graph with 10 inputs, 25 function nodes, and 10 outputs.

simulate_3_in_3_out(n_train, n_test[, ...])

simulate_sde(G, n_train, n_test, ...[, ...])

Generate samples from a synthetic graph-structured data-generation process using stochastic ODEs.
gsnn.gsnn.simulate.datasets.simulate_10_in_25_func_10_out_cyclic(n_train, n_test, noise_scale=0.1, device='cpu', zscorey=False, dt=0.01, t_final=10.0, seed=None)[source]

Create a complex cyclic graph with 10 inputs, 25 function nodes, and 10 outputs. Maximum path length from input to output is 10. Uses SDE method for data generation.

Parameters:

  • n_train (int) – Number of training samples

  • n_test (int) – Number of test samples

  • noise_scale (float) – Noise scale for SDE integration

  • device (str) – Device to place tensors on

  • zscorey (bool) – Whether to z-score normalize y values

  • dt (float) – Time step for SDE integration

  • t_final (float) – Final time for SDE integration

  • seed (int) – Random seed for reproducibility

Returns:

Tuple containing graph, positions, train/test data, and node lists

gsnn.gsnn.simulate.datasets.simulate_3_in_3_out(n_train, n_test, noise_scale=0.1, device='cpu', zscorey=False)[source]