Datasets:
File size: 2,343 Bytes
6c1c833 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 | {
"field_id": "F028",
"field": "causal discovery",
"directions": [
{
"direction_id": "G0",
"name": "Causal structure recovery for specified variables",
"scope": "Infer causal relationships among variables from observational data and any available external structural knowledge. This includes pairwise causal direction, multivariable graphs and distributions over candidate graphs, as well as structural identification problems arising from measurement error, latent confounding, proxy variables, or hidden nodes."
},
{
"direction_id": "G1",
"name": "Identification of latent causal representations",
"scope": "When observations do not directly provide the target causal variables, identify or learn latent causal variables and their relationships, potentially using interactions, interventions, environmental variation, or observational structure."
},
{
"direction_id": "G2",
"name": "Temporal causal structure discovery",
"scope": "Identify directed dependencies or causal mechanisms in evolving processes, including lagged relationships, contemporaneous effects, events, and dynamical systems."
},
{
"direction_id": "G3",
"name": "Causal structure discovery across contexts and regimes",
"scope": "Use or characterize relationships across populations, conditions, studies, temporal scales, and intervention environments to discover causal structure. This includes shared backbones and differing structures, as well as identifying relationships among variables through mechanism changes or invariance."
},
{
"direction_id": "G4",
"name": "Active causal experimentation",
"scope": "Select informative actions or interventions to learn causal relationships, resolve graph uncertainty, or answer a target causal query, while considering experimental efficiency or cost."
},
{
"direction_id": "G5",
"name": "Validation and evaluation of causal discovery",
"scope": "Assess the credibility of causal-structure claims or discovery procedures by testing, falsifying, or auditing candidate relationships and graphs, and by developing test materials and systematically evaluating discovery methods for graph recovery, downstream utility, or sensitivity to methodological choices."
}
]
}
|