fastLoess R API Reference¶

The R bindings provide a high-performance interface to the core Rust library, mirroring the Rust API structure.

Classes¶

`Loess`¶

The Loess class allows configuring the LOESS parameters once and fitting multiple datasets using those parameters.

Constructor:

model <- Loess(fraction = 0.5)

...: Arguments corresponding to LoessOptions fields.

Methods:

model <- Loess()
result <- model$fit(x, y)
# or with per-observation weights:
result <- model$fit(x, y, custom_weights = weights)

Fits the model to the provided x and y numeric vectors.
Returns a LoessResult S3 object containing the smoothed values and optional diagnostics.
print(model): Displays the model configuration.

`StreamingLoess`¶

The StreamingLoess class processes data in chunks, suitable for very large datasets or streaming applications.

Constructor:

stream <- StreamingLoess(chunk_size = 10L)

...: Arguments corresponding to LoessOptions and StreamingOptions fields.

Methods:

stream <- StreamingLoess(chunk_size = 10L)
partial_result <- stream$process_chunk(x, y)

Processes a chunk of data. Returns partial results.

stream <- StreamingLoess(chunk_size = 10L)
stream$process_chunk(x, y)
final_result <- stream$finalize()

Finalizes the smoothing process and returns any remaining buffered results.

`OnlineLoess`¶

The OnlineLoess class updates the model incrementally with new data points.

Constructor:

online <- OnlineLoess(window_capacity = 10L)

...: Arguments corresponding to LoessOptions and OnlineOptions fields.

Methods:

online <- OnlineLoess(window_capacity = 10L)
result <- online$add_point(x[[1L]], y[[1L]])  # returns list or NULL

Adds a single point to the sliding window. Returns a named list ($smoothed, $residual, …) once the window has enough points, or NULL while still filling.

Options Structures¶

`LoessOptions`¶

Field	Type	Default	Description
`fraction`	`numeric`	`0.67`	Smoothing fraction (bandwidth)
`iterations`	`integer`	`3`	Number of robustifying iterations
`weight_function`	`character`	`"tricube"`	Kernel weight function
`robustness_method`	`character`	`"bisquare"`	Robustness method
`scaling_method`	`character`	`"mad"`	Residual scaling method
`boundary_policy`	`character`	`"extend"`	Boundary handling policy
`zero_weight_fallback`	`character`	`"use_local_mean"`	Zero-weight handling
`auto_converge`	`numeric`	`NULL`	Auto-convergence tolerance
`custom_weights`	`numeric`	`NULL`	Per-observation case weights — passed to `$fit()`, not the constructor (Batch only)
`confidence_intervals`	`numeric`	`NULL`	Confidence level (e.g., 0.95)
`prediction_intervals`	`numeric`	`NULL`	Prediction level (e.g., 0.95)
`return_diagnostics`	`logical`	`FALSE`	Compute RMSE, MAE, R², AIC
`return_residuals`	`logical`	`FALSE`	Include residuals in result
`return_robustness_weights`	`logical`	`FALSE`	Include robustness weights in result
`return_se`	`logical`	`FALSE`	Compute hat-matrix statistics (enp, leverage …)
`parallel`	`logical`	`TRUE`	Enable parallel execution
`degree`	`character`	`"linear"`	Polynomial degree of local fit
`dimensions`	`integer`	`1L`	Number of predictor dimensions
`distance_metric`	`character`	`"normalized"`	Distance metric; use `"minkowski:p"` for custom p
`weighted_metric_weights`	`numeric`	`NULL`	Per-dimension weights (used when `distance_metric = "weighted"`)
`surface_mode`	`character`	`"interpolation"`	Surface computation mode
`cell`	`numeric`	`NULL`	Cell size for interpolation grid (smaller → more vertices, higher accuracy)
`interpolation_vertices`	`integer`	`NULL`	Number of interpolation vertices
`boundary_degree_fallback`	`logical`	`NULL`	Fall back to lower polynomial degree at boundaries when higher degrees fail
`cv_method`	`character`	`"kfold"`	CV method (`"kfold"` or `"loocv"`) (Batch only)
`cv_k`	`integer`	`5L`	Number of folds for k-fold CV (Batch only)
`cv_fractions`	`numeric`	`NULL`	Fractions to test for cross-validation (Batch only)
`cv_seed`	`integer`	`NULL`	Random seed for cross-validation shuffling (Batch only)

`StreamingOptions` (inherits `LoessOptions`)¶

Field	Type	Default	Description
`chunk_size`	`integer`	`5000L`	Data chunk size
`overlap`	`integer`	`500L`	Overlap between chunks
`merge_strategy`	`character`	`"weighted_average"`	Strategy for blending overlap regions

`OnlineOptions` (inherits `LoessOptions`)¶

Field	Type	Default	Description
`window_capacity`	`integer`	`1000L`	Max points in sliding window
`min_points`	`integer`	`3L`	Min points before smoothing starts
`update_mode`	`character`	`"full"`	Update mode (`"full"` or `"incremental"`)
`parallel`	`logical`	`FALSE`	Enable parallel execution (off by default; online LOESS fits one point at a time)

Result Structure¶

`OnlineOutput` (named list)¶

Returned by add_point() once the window has enough points (NULL until then).

Field	Type	Description
`smoothed`	`numeric`	Smoothed value for the latest point
`std_error`	`numeric` (optional)	Standard error (if requested)
`residual`	`numeric` (optional)	Residual y − smoothed (if requested)
`robustness_weight`	`numeric` (optional)	Robustness weight (if requested)
`iterations_used`	`integer` (optional)	Robustness iterations performed

`LoessResult`¶

An S3 list with class "LoessResult" containing:

Supported S3 Methods: print(result), plot(result)

Field	Type	Description
`x`	`numeric`	Sorted x values
`y`	`numeric`	Smoothed y values
`fraction_used`	`numeric`	Fraction used (set or selected by CV)
`iterations_used`	`integer \\| NULL`	Robustness iterations actually performed
`standard_errors`	`numeric \\| NULL`	Per-point SE (if `return_se`)
`confidence_lower`	`numeric \\| NULL`	Lower confidence bounds
`confidence_upper`	`numeric \\| NULL`	Upper confidence bounds
`prediction_lower`	`numeric \\| NULL`	Lower prediction bounds
`prediction_upper`	`numeric \\| NULL`	Upper prediction bounds
`residuals`	`numeric \\| NULL`	Residuals (if `return_residuals`)
`robustness_weights`	`numeric \\| NULL`	Robustness weights (if `return_robustness_weights`)
`cv_scores`	`numeric \\| NULL`	CV score per tested fraction
`diagnostics`	`list \\| NULL`	Fit metrics (if `return_diagnostics`)
`enp`	`numeric \\| NULL`	Equivalent number of parameters (if `return_se`)
`trace_hat`	`numeric \\| NULL`	Trace of hat matrix (if `return_se`)
`delta1`	`numeric \\| NULL`	First delta statistic (if `return_se`)
`delta2`	`numeric \\| NULL`	Second delta statistic (if `return_se`)
`residual_scale`	`numeric \\| NULL`	Residual scale estimate (if `return_se`)
`leverage`	`numeric \\| NULL`	Per-point hat-matrix diagonal (if `return_se`)
`dimensions`	`integer`	Number of predictor dimensions

`Diagnostics`¶

Field	Type	Description
`rmse`	`numeric`	Root Mean Squared Error
`mae`	`numeric`	Mean Absolute Error
`r_squared`	`numeric`	R-squared
`residual_sd`	`numeric`	Residual standard deviation
`effective_df`	`numeric`	Effective degrees of freedom (NaN if not computed)
`aic`	`numeric`	AIC (NaN if not computed)
`aicc`	`numeric`	AICc (NaN if not computed)

Options¶

weight_function¶

"tricube" (default)
"epanechnikov"
"gaussian"
"uniform" (alias: "boxcar")
"biweight" (alias: "bisquare")
"triangle" (alias: "triangular")
"cosine"

robustness_method¶

"bisquare" (default; alias: "biweight")
"huber"
"talwar"

boundary_policy¶

"extend" (default; alias: "pad")
"reflect" (alias: "mirror")
"zero"
"noboundary" (alias: "none")

scaling_method¶

"mad" (default; alias: "median_absolute_deviation")
"mar" (alias: "median_absolute_residual")
"mean" (alias: "mean_absolute_residual")

zero_weight_fallback¶

"use_local_mean" (default; aliases: "local_mean", "mean")
"return_original" (alias: "original")
"return_none" (alias: "none")

degree¶

"constant" or "0" (degree 0)
"linear" or "1" (default, degree 1)
"quadratic" or "2" (degree 2)
"cubic" or "3" (degree 3)
"quartic" or "4" (degree 4)

distance_metric¶

"normalized" (default — scales each dimension by its range; alias: "norm")
"euclidean" (alias: "euclid")
"manhattan" (alias: "l1")
"chebyshev" (alias: "linf")
"minkowski" (Euclidean when no suffix; use "minkowski:p" for custom p, e.g. "minkowski:3")
"weighted" plus weighted_metric_weights for per-dimension scaling (alias: "weighted_euclidean")

surface_mode¶

"interpolation" (default — faster, uses a spatial grid)
"direct" (fits every point exactly; slower but more accurate)

merge_strategy¶

"weighted_average" (default; alias: "weighted")
"average" (alias: "mean")
"take_first" (alias: "first")
"take_last" (alias: "last")

update_mode¶

"full" (default; alias: "resmooth")
"incremental" (alias: "single")

Example¶

library(rfastloess)

x <- seq(0, 10, length.out = 100)
y <- sin(x) + rnorm(100, sd = 0.2)

# Configure model
model <- Loess(fraction = 0.5)

# Fit data
result <- model$fit(x, y)

# Print summary
print(result)

# Plot result
plot(result)

fastLoess R API Reference¶

Classes¶

Loess¶

StreamingLoess¶

OnlineLoess¶

Options Structures¶

LoessOptions¶

StreamingOptions (inherits LoessOptions)¶

OnlineOptions (inherits LoessOptions)¶

Result Structure¶

OnlineOutput (named list)¶

LoessResult¶

Diagnostics¶

Options¶

weight_function¶

robustness_method¶

boundary_policy¶

scaling_method¶

zero_weight_fallback¶

degree¶

distance_metric¶

surface_mode¶

merge_strategy¶

update_mode¶

Example¶

`Loess`¶

`StreamingLoess`¶

`OnlineLoess`¶

`LoessOptions`¶

`StreamingOptions` (inherits `LoessOptions`)¶

`OnlineOptions` (inherits `LoessOptions`)¶

`OnlineOutput` (named list)¶

`LoessResult`¶

`Diagnostics`¶