fastLoess & loess-rs Rust API Reference¶

The Rust crates provide the core implementation and high-performance extensions.

Structs & Usage¶

Both crates expose the same three entry types via their prelude: Loess for batch mode, StreamingLoess for chunked processing, and OnlineLoess for sliding-window updates.

use fastLoess::prelude::*;  // or: use loess_rs::prelude::*;

`Loess` (Batch)¶

Standard in-memory smoothing.

Constructor:

let model = Loess::new().build()?;

Methods:

let model = Loess::new().build()?;
let result = model.fit(&x, &y)?;

Fits the model to the provided x and y arrays.
Returns Result<LoessResult<T>, LoessError>.

`StreamingLoess`¶

Streaming mode for large datasets.

Constructor:

let mut processor = StreamingLoess::new().build()?;

Methods:

let mut processor = StreamingLoess::new().build()?;
let result = processor.process_chunk(&x, &y)?;

Processes a chunk of data. Returns LoessResult<T> with partial results.

let mut processor = StreamingLoess::new().build()?;
processor.process_chunk(&x_chunk, &y_chunk)?;
let final_result = processor.finalize()?;

Finalizes processing and returns remaining buffered results.

`OnlineLoess`¶

Online mode for real-time data.

Constructor:

let mut processor = OnlineLoess::new().build()?;

Methods:

let mut processor = OnlineLoess::new().build()?;
let output = processor.add_point(&[x[0]], y[0])?;

Adds a single point (x, y) to the window.
Returns Result<Option<OnlineOutput<T>>, LoessError>.

let mut processor = OnlineLoess::new().build()?;
processor.reset();

Clears the internal window buffer.

Builder Configuration¶

These chained methods configure the builder. They correspond to the "Options Structures" in other bindings.

Loess Options¶

Method	Default	Description
`fraction(T)`	`0.67`	Smoothing fraction (bandwidth)
`iterations(usize)`	`3`	Number of robustifying iterations
`weight_function(...)`	`"tricube"`	Kernel weight function
`robustness_method(...)`	`"bisquare"`	Robustness method
`scaling_method(...)`	`"mad"`	Residual scaling method
`boundary_policy(...)`	`"extend"`	Boundary handling policy
`zero_weight_fallback(...)`	`"use_local_mean"`	Zero-weight handling
`auto_converge(T)`	disabled	Auto-convergence tolerance
`custom_weights(Vec<T>)`	disabled	Per-observation case weights (Batch only)
`confidence_intervals(T)`	disabled	Confidence level (e.g., 0.95)
`prediction_intervals(T)`	disabled	Prediction level (e.g., 0.95)
`return_diagnostics()`	`false`	Compute RMSE, MAE, R², AIC
`return_residuals()`	`false`	Include residuals in result
`return_robustness_weights()`	`false`	Include robustness weights in result
`return_se()`	`false`	Compute hat-matrix statistics (enp, leverage …)
`parallel(bool)`	`true`	Enable parallel execution
`degree(...)`	`"linear"`	Polynomial degree
`dimensions(usize)`	`1`	Number of predictor dimensions
`distance_metric(...)`	`"normalized"`	Distance metric
`weighted_metric_weights(Vec<T>)`	disabled	Per-dimension weights (used when `distance_metric = "weighted"`)
`surface_mode(...)`	`"interpolation"`	Surface computation mode
`cell(T)`	disabled	Cell size for interpolation grid (smaller → more vertices, higher accuracy)
`interpolation_vertices(usize)`	disabled	Number of interpolation vertices
`boundary_degree_fallback(bool)`	`true`	Fall back to lower polynomial degree at boundaries when higher degrees fail
`cv_method(...)`	disabled	Cross-validation method
`cv_k(...)`	disabled	Number of folds for K-fold cross-validation
`cv_fractions(...)`	disabled	Candidate fractions to evaluate during cross-validation
`cv_seed(...)`	disabled	Random seed for reproducible fold assignments

Streaming Options¶

Method	Default	Description
`chunk_size(usize)`	`5000`	Data chunk size
`overlap(usize)`	`500`	Overlap between chunks
`merge_strategy(...)`	`"weighted_average"`	Strategy for blending overlap regions

Online Options¶

Method	Default	Description
`window_capacity(usize)`	`1000`	Max points in sliding window
`min_points(usize)`	`3`	Min points before smoothing starts
`update_mode(...)`	`"full"`	Update strategy
`parallel(bool)`	`false`	Enable parallel execution (off by default; online LOESS fits one point at a time)

Result Structure¶

`LoessResult<T>`¶

Field	Type	Description
`x`	`Vec<T>`	Sorted x values
`y`	`Vec<T>`	Smoothed y values
`fraction_used`	`T`	Fraction used (set or selected by CV)
`iterations_used`	`Option<usize>`	Robustness iterations actually performed
`standard_errors`	`Option<Vec<T>>`	Per-point SE (if `return_se()`)
`confidence_lower`	`Option<Vec<T>>`	Lower confidence bounds
`confidence_upper`	`Option<Vec<T>>`	Upper confidence bounds
`prediction_lower`	`Option<Vec<T>>`	Lower prediction bounds
`prediction_upper`	`Option<Vec<T>>`	Upper prediction bounds
`residuals`	`Option<Vec<T>>`	Residuals (if `return_residuals()`)
`robustness_weights`	`Option<Vec<T>>`	Robustness weights (if `return_robustness_weights()`)
`cv_scores`	`Option<Vec<T>>`	CV score per tested fraction
`diagnostics`	`Option<Diagnostics<T>>`	Fit metrics (if `return_diagnostics()`)
`enp`	`Option<T>`	Equivalent number of parameters (if `return_se()`)
`trace_hat`	`Option<T>`	Trace of hat matrix (if `return_se()`)
`delta1`	`Option<T>`	First delta statistic (if `return_se()`)
`delta2`	`Option<T>`	Second delta statistic (if `return_se()`)
`residual_scale`	`Option<T>`	Residual scale estimate (if `return_se()`)
`leverage`	`Option<Vec<T>>`	Per-point hat-matrix diagonal (if `return_se()`)
`dimensions`	`usize`	Number of predictor dimensions
`polynomial_degree`	`PolynomialDegree` (internal)	Polynomial degree used; implements `Display` (e.g. `"linear"`)
`distance_metric`	`DistanceMetric<T>` (internal)	Distance metric used; implements `Display` (e.g. `"normalized"`)

`Diagnostics<T>`¶

Field	Type	Description
`rmse`	`T`	Root Mean Squared Error
`mae`	`T`	Mean Absolute Error
`r_squared`	`T`	R-squared
`residual_sd`	`T`	Residual standard deviation
`effective_df`	`Option<T>`	Effective degrees of freedom
`aic`	`Option<T>`	AIC
`aicc`	`Option<T>`	AICc

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")

distance_metric¶

"normalized" (default — scales each dimension by its range; alias: "norm")
"euclidean" (alias: "euclid")
"manhattan" (alias: "l1")
"chebyshev" (alias: "linf")
"minkowski" or "minkowski:p" for a custom exponent
"weighted" plus .weighted_metric_weights(vec![...]) (alias: "weighted_euclidean")

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)

surface_mode¶

"interpolation" (default)
"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¶

let model = Loess::new()
    .fraction(0.5)
    .iterations(3)
    .build()?;

let result = model.fit(&x, &y)?;

println!("Smoothed Y: {:?}", result.y);

fastLoess & loess-rs Rust API Reference¶

Structs & Usage¶

Loess (Batch)¶

StreamingLoess¶

OnlineLoess¶

Builder Configuration¶

Loess Options¶

Streaming Options¶

Online Options¶

Result Structure¶

LoessResult<T>¶

Diagnostics<T>¶

Options¶

weight_function¶

robustness_method¶

boundary_policy¶

scaling_method¶

zero_weight_fallback¶

distance_metric¶

degree¶

surface_mode¶

merge_strategy¶

update_mode¶

Example¶

`Loess` (Batch)¶

`StreamingLoess`¶

`OnlineLoess`¶

`LoessResult<T>`¶

`Diagnostics<T>`¶