ml-spectra-processing
    Preparing search index...

    Interface XWhittakerSmootherOptions

    interface XWhittakerSmootherOptions {
        algorithm?: "cholesky" | "thomas";
        controlPoints?: NumberArray;
        factorStd?: number;
        lambda?: number;
        learningRate?: number;
        maxIterations?: number;
        tolerance?: number;
        weights?: NumberArray;
    }

    Hierarchy (View Summary)

    Index

    Properties

    algorithm?: "cholesky" | "thomas"

    Solver algorithm to use for the smoothing routine.

    • 'cholesky' — build the full (symmetric) system and solve using a Cholesky decomposition. Generally more robust for small/medium arrays and when a matrix factorization is acceptable.
    • 'thomas' — use a tridiagonal formulation and solve with the Thomas algorithm (specialised tridiagonal solver). Lower memory usage and faster for large inputs when the system is tridiagonal.
    'thomas'
    
    controlPoints?: NumberArray

    Control points to determine which weights to update.

    factorStd?: number

    Factor used to calculate the threshold for determining outliers in the residuals. Higher values mean more tolerance for outliers. The default value is based on noise follow the normal distribution values over 3 times the standard-deviation could be marked as signals or outliers.

    3
    
    lambda?: number

    Factor of weights matrix in -> [I + lambda D'D]z = x

    100
    
    learningRate?: number

    Learning rate for weight updates.

    0.5
    
    maxIterations?: number

    Maximum number of iterations for the baseline refinement process.

    100
    
    tolerance?: number

    Tolerance for convergence. The process stops if the change in baseline is less than this value.

    1e-6
    
    weights?: NumberArray

    Array of weights

    [1,1,...,1]