AutoMovie API
    Preparing search index...

    Interface IAutoMovieRecoveryMapJointFit

    Local joint-refinement diagnostics on one fixed observation population.

    interface IAutoMovieRecoveryMapJointFit {
        acceptedSteps: number;
        converged: boolean;
        curvature?: IAutoMovieRecoveryCurvatureReport;
        feasible?: IAutoMovieRecoveryFeasibleDiagnostics;
        finalCost: number;
        finalSquaredError: number;
        initialCost: number;
        initialSquaredError: number;
        iterations: number;
        minimumCameraDesignRatio: number;
        minimumPointDesignRatio: number;
        observationCount: number;
        regularization?: IAutoMovieRecoveryRegularizationReport;
        termination:
            | "step"
            | "iteration-budget"
            | "gradient"
            | "cost"
            | "direction-incomplete"
            | "abnormal-or-stalled"
            | "pixel-kkt";
    }
    Index
    acceptedSteps: number

    Selected state updates decreasing the actual complete-population robust objective; qualified unselected evaluations are excluded.

    converged: boolean

    Local numerical stopping condition, not global accuracy or uncertainty.

    Present only for requested internal lagged curvature; fixed model-use counters are independent of original rank, actual selection and normal convergence.

    Completed feasible-fit normal/QP/SQP/trial diagnostics; absent for ordinary fits and older artifacts, without implying no fit ran. Nonfinite values use explicit JSON sentinels.

    finalCost: number

    Final robust sum over exactly observationCount measurements, in squared pixels.

    finalSquaredError: number

    Final unweighted squared error; robust descent need not decrease this distinct quantity.

    initialCost: number

    Initial sum s^2*(sqrt(1+norm2(r)/s^2)-1), in squared pixels.

    initialSquaredError: number

    Initial sum of squared two-coordinate pixel residual norms over the same observations.

    iterations: number

    Nonlinear attempts; feasible mode counts every actual evaluation, including qualified unselected and rejected candidates.

    minimumCameraDesignRatio: number

    Smallest observed undamped column-scaled reduced-camera singular ratio.

    minimumPointDesignRatio: number

    Smallest observed undamped column-scaled point-block singular ratio.

    observationCount: number

    Same observations used in both initial and final objectives; none is silently dropped during this fit.

    Present only for internal accepted-agreement regularization; fit-local scalar reactions never replace normal convergence or data rank.

    termination:
        | "step"
        | "iteration-budget"
        | "gradient"
        | "cost"
        | "direction-incomplete"
        | "abnormal-or-stalled"
        | "pixel-kkt"

    Numerical stop; pixel-kkt is normal constrained stationarity, while direction-incomplete, abnormal-or-stalled and iteration-budget are nonconverged.