LOWESS Online Smoothing — OnlineLowess • rfastlowess

Create a stateful LOWESS model for real-time online data. Maintains a sliding window and processes each incoming point immediately via add_point.

Usage

OnlineLowess(
    fraction = 0.67,
    ...,
    iterations = 0L,
    weight_function = "tricube",
    robustness_method = "bisquare",
    delta = NULL,
    zero_weight_fallback = "use_local_mean",
    boundary_policy = "extend",
    scaling_method = "mad",
    auto_converge = NULL,
    missing = "error",
    window_capacity = 1000L,
    min_points = 2L,
    update_mode = "incremental",
    outputs = NULL,
    intervals = NULL,
    seed = NULL
)

Arguments

fraction: Smoothing fraction, greater than 0 and up to 1. Default: 0.67.
...: Not used; forces all subsequent arguments to be named.
iterations: Number of robustness iterations. Requires update_mode = "full"; the default "incremental" mode is a non-robust single-point fit that ignores robustness iterations. Default: 0.
weight_function: Kernel weight function. One of "tricube" (default), "gaussian", "uniform" (alias: "boxcar"), "cosine", "epanechnikov", "biweight" (alias: "bisquare"), or "triangle" (alias: "triangular").
robustness_method: Outlier downweighting method: "bisquare" (default; alias: "biweight"), "huber", or "talwar".
delta: Interpolation distance as a non-negative fraction of the x range. In Online mode, NULL (default) disables interpolation; positive values require update_mode = "full".
zero_weight_fallback: Fallback policy when all robustness weights drop to zero: "use_local_mean" (default; aliases: "local_mean", "mean"), "return_original" (alias: "original"), or "return_none" (alias: "none").
boundary_policy: Boundary handling strategy: "extend" (default; alias: "pad"), "reflect" (alias: "mirror"), "zero", or "noboundary" (alias: "none").
scaling_method: Residual scale estimation for robustness weights: "mad" (default; alias: "median_absolute_deviation"), "mar" (alias: "median_absolute_residual"), or "mean" (alias: "mean_absolute_residual").
auto_converge: Convergence tolerance for early stopping. Requires update_mode = "full" and iterations > 0; NULL (default) disables early stopping.
missing: Policy for non-finite (NaN/Infinity) values in input data: "error" (default) raises an error, "drop" silently removes affected observations before fitting.
window_capacity: Maximum number of points kept in the sliding window, at least 3. Default: 1000.
min_points: Minimum number of points required before smoothing begins, between 2 and window_capacity. Default: 2.
update_mode: Window update strategy: "incremental" (default; alias: "single") updates only the newest point and does not run robustness iterations; "full" (alias: "resmooth") re-smooths all window points after each addition and supports robustness iterations, positive delta, and auto_converge.
outputs: Character vector selecting optional Online output components: "se" (standard errors), "weights" (robustness weights), and/or "derivative". NULL (default) returns only the core result.
intervals: Interval options, created with intervals_opts (or a named list with any of confidence, prediction, bootstrap): e.g. intervals = intervals_opts(confidence = 0.90, prediction = 0.99, bootstrap = 200). Confidence and prediction coverage levels are independent and may differ. NULL (default) disables intervals.
seed: Non-negative whole-number seed shared by cross-validation and bootstrap resampling for reproducible results. When cross-validation is unavailable, it controls bootstrap resampling only. NULL (default) uses a random seed.

Value

An OnlineLowess object.

Details

Best suited when data arrives incrementally (e.g. sensors or streams), real-time smoothed values are needed, or memory is fixed. For datasets that fit in memory, see Lowess; for large batches processed in chunks, see StreamingLowess.

Confidence and prediction interval coverage levels are independent; both interval types require update_mode = "full". Positive delta values also require update_mode = "full". A non-NULL auto_converge requires full mode and at least one robustness iteration; unsupported combinations are rejected.

Examples

model <- OnlineLowess(fraction = 0.2, window_capacity = 20)
x <- 1:50
y <- sin(x * 0.1) + rnorm(50, 0, 0.1)
smoothed <- numeric(0)
for (i in seq_along(x)) {
    result <- add_point(model, x[i], y[i])
    if (!is.null(result)) smoothed <- c(smoothed, result$y)
}
head(smoothed, 5)
#> [1] 0.1201261 0.1898465 0.3098642 0.3037980 0.4955887