GAE
Package | gCastle |
Version | 1.0.3 |
Language | Python |
Docs | |
Paper | :footcite:t:`https://doi.org/10.48550/arxiv.1911.07420` |
Graph type | DAG |
Docker | bpimages/gcastle:1.0.3 |
Module folder | gcastle_gae |
Description
A gradient-based algorithm using graph autoencoder to model non-linear causal relationships.
Example JSON
[
{
"id": "gcastle_gae",
"x_dim": 1,
"num_encoder_layers": 1,
"num_decoder_layers": 1,
"hidden_size": 4,
"latent_dim": 1,
"l1_graph_penalty": 0.0,
"use_float64": false,
"learning_rate": "1e-3",
"max_iter": 10,
"iter_step": 3000,
"init_iter": 3,
"h_tol": "1e-8",
"init_rho": 1.0,
"rho_thres": "1e+30",
"h_thres": 0.25,
"rho_multiply": 2.0,
"early_stopping": false,
"early_stopping_thres": 1.0,
"graph_thres": 0.3,
"timeout": null
}
]
.. footbibliography::