ANMNonlinear (gCastle)
Module name |
|
Package |
|
Version |
1.0.3 |
Language |
|
Docs |
|
Paper |
Hoyer et al.[1] |
Graph type |
|
MCMC |
No |
Edge constraints |
No |
Data type |
C, D |
Data missingness |
|
Intervention type |
|
Docker |
ANMNonlinear
Nonlinear causal discovery with additive noise models.
Example
Config file: gcastle_nonlinear.json
Command:
snakemake --cores all --use-apptainer --configfile config/gcastle_nonlinear.json
Fig. 48 and Fig. 49 show the pattern graph’s FP/P vs. TP/P benchmark results for gCastle algorithms, and comparison with BOSS (TETRAD) and Iterative MCMC (BiDAG), tested on nonlinear data. The benchmark is based on 5 datasets corresponding to 5 realisations of a 20-variable random nonlinear Gaussian SEM with Erdős-Rényi structure (expected degree 4, max parents 5). The nonlinear relationships are modeled using multi-layer perceptrons (MLP) and quadratic functions, see gcastle_iidsim. Each dataset contains 300 standardized samples. The SEM parameters are uniformly sampled from [0.25, 1].
Fig. 24 FP/P vs. TP/P for gCastle algorithms on nonlinear MLP data.
Fig. 25 FP/P vs. TP/P for gCastle algorithms on nonlinear quadratic data.
Example JSON
[
{
"id": "gcastle_anm",
"alpha": 0.05,
"timeout": null
}
]