Member Junction
    Preparing search index...

    Line 1 of the matrix file: everything the trainer needs to read the rows that follow.

    interface SnapshotMatrixHeader {
        columns: readonly string[];
        decision: DecisionSpec;
        featureColumns: readonly SnapshotFeatureColumnHeader[];
        featureSchema: readonly FeatureSchemaEntry[];
        format: "ps-snapshot-matrix/v1";
        incumbentScoreColumn: "__ps_incumbent_score" | null;
        labelColumn: string;
        preprocessing: readonly PreprocessingOp[];
        referenceScoreColumn: "__ps_reference_score" | null;
        sliceColumns: Readonly<Record<string, string>>;
    }
    Index

    Properties

    columns: readonly string[]

    Every column, in row-cell order.

    decision: DecisionSpec
    featureColumns: readonly SnapshotFeatureColumnHeader[]
    featureSchema: readonly FeatureSchemaEntry[]

    The sidecar feature schema: the run's FeaturesUsed and the model's FeatureSchema record it.

    format: "ps-snapshot-matrix/v1"
    incumbentScoreColumn: "__ps_incumbent_score" | null
    labelColumn: string
    preprocessing: readonly PreprocessingOp[]
    referenceScoreColumn: "__ps_reference_score" | null
    sliceColumns: Readonly<Record<string, string>>

    Slice field → its __ps_slice:<field> column.