causalcompass.algorithms.NTSNotears

class causalcompass.algorithms.NTSNotears(tau_max=3, wthre=0.01, lambda_1=0.001, lambda_2=0.01, device='cuda', max_iter=100, h_tol=1e-8, rho_max=1e+16, seed=None)[source]

NTS-NOTEARS is a score-based nonlinear extension of DYNOTEARS.

References

https://github.com/xiangyu-sun-789/NTS-NOTEARS

Parameters:

  • tau_max (int, default 3) – Maximum time lag

  • wthre (float or list, default 0.01) – Weight threshold for edge filtering

  • lambda_1 (float, list, or str, default 0.001) – L1 regularization parameter

  • lambda_2 (float, default 0.01) – L2 regularization parameter

  • device (str, default 'cuda') – Computation device

  • max_iter (int, default 100) – Max number of dual ascent steps during optimization

  • h_tol (float, default 1e-8) – Tolerance for acyclicity constraint

  • rho_max (float, default 1e+16) – Maximum value for the augmented Lagrangian penalty parameter

Examples

>>> from causalcompass.algorithms import NTSNotears
>>> model = NTSNotears(tau_max=3, wthre=0.01, lambda_1=0.001, lambda_2=0.01, device='cuda')
>>> predicted_adj = model.run(X)
>>> all_metrics, no_diag_metrics = model.eval(true_adj, predicted_adj)
__init__(tau_max=3, wthre=0.01, lambda_1=0.001, lambda_2=0.01, device='cuda', max_iter=100, h_tol=1e-8, rho_max=1e+16, seed=None)[source]

Initialize NTS-NOTEARS

Methods

__init__([tau_max, wthre, lambda_1, ...])

Initialize NTS-NOTEARS

eval(true_adj, predicted_adj[, shd_thresholds])

Evaluate the predicted adjacency matrix against the ground truth.

run(X)

Run NTS-NOTEARS algorithm.

run_raw(X)

Run the algorithm and return an unthresholded intermediate result that can be reused across multiple threshold values.

run_threshold_sweep(X, thresholds)

Run the algorithm once and post-process the raw result for each threshold.