Class StructuralCausalModel

Index

Constructors

Accessors

Methods

abduct anomalyScores attributeAnomalies counterfactual detectDistributionChange train

Constructors

constructor

new StructuralCausalModel(graph: CausalGraph): StructuralCausalModel
Parameters
- graph: CausalGraph
Returns StructuralCausalModel
- Defined in pipeline/src/gcm/structural-causal-model.ts:57

Accessors

causalGraph

get causalGraph(): CausalGraph
Returns CausalGraph
- Defined in pipeline/src/gcm/structural-causal-model.ts:219

Methods

abduct

abduct(observation: Record<string, number>): Record<string, number>
Abduction: infer noise from observation
Parameters
- observation: Record<string, number>
Returns Record<string, number>
- Defined in pipeline/src/gcm/structural-causal-model.ts:132

anomalyScores

anomalyScores(observation: Record<string, number>): Map<string, number>
Score each node by how anomalous its current value is under the SCM
Parameters
- observation: Record<string, number>
Returns Map<string, number>
- Defined in pipeline/src/gcm/structural-causal-model.ts:162

attributeAnomalies

attributeAnomalies(
observation: Record<string, number>,
topK?: number,
): RootCause[]
Attribute anomalies: find root nodes with highest anomaly scores
Parameters
- observation: Record<string, number>
- topK: number = 3
Returns RootCause[]
- Defined in pipeline/src/gcm/structural-causal-model.ts:177

counterfactual

counterfactual(
noise: Record<string, number>,
intervention: Record<string, number>,
): Record<string, number>
Action + Prediction: apply intervention and forward-simulate
Parameters
- noise: Record<string, number>
- intervention: Record<string, number>
Returns Record<string, number>
- Defined in pipeline/src/gcm/structural-causal-model.ts:144

detectDistributionChange

detectDistributionChange(
before: Record<string, number>[],
after: Record<string, number>[],
): { changed: boolean; meanShift: number; pValue: number }
Detect if the observation's anomaly pattern indicates distribution change
Parameters
- before: Record<string, number>[]
- after: Record<string, number>[]
Returns { changed: boolean; meanShift: number; pValue: number }
- Defined in pipeline/src/gcm/structural-causal-model.ts:194

train

train(data: number[][]): void
Train mechanisms from data: X_i = f(pa(X_i)) + ε
Parameters
- data: number[][]
Returns void
- Defined in pipeline/src/gcm/structural-causal-model.ts:63