fastLoess WebAssembly API Reference

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

StreamingLoess and OnlineLoess are documented separately: Streaming Adapter, Online Adapter

When to Use Batch Adapter

• Dataset fits in memory
• Need intervals, cross-validation, or diagnostics
• Processing complete files

Classes

Loess

The Loess class is the main entry point for batch smoothing.

Constructor:

const { Loess } = require('fastloess-wasm');

const model = new Loess({ fraction: 0.5 });
console.log("typeof fit:", typeof model.fit);

typeof fit: function

• options: An object containing LoessOptions fields.

fit(x, y)

Fits the model to the provided x and y typed arrays. Returns a LoessResult object.

const { Loess } = require('fastloess-wasm');

const n = 100;
const x = Float64Array.from({ length: n }, (_, i) => i * 2 * Math.PI / (n - 1));
const y = Float64Array.from(x, xi => Math.sin(xi) + 0.1);

const model = new Loess({ fraction: 0.5 });
const result = model.fit(x, y);
console.log("Fraction used:", result.fraction_used);
console.log("Iterations used:", result.iterations_used);
// or with per-observation weights:
const weights = new Float64Array(n).fill(1);
const resultWeighted = model.fit(x, y, weights);

Fraction used: 0.5
Iterations used: 3

Options Structures

LoessOptions

Field	Type	Default	Description
fraction	number	0.67	Smoothing fraction (bandwidth)
iterations	number	3	Number of robustifying iterations
weight_function	string	"tricube"	Kernel weight function
robustness_method	string	"bisquare"	Robustness method
degree	string	"linear"	Polynomial degree of local fit
dimensions	number	1	Number of predictor dimensions
distance_metric	string	"normalized"	Distance metric; use "minkowski:p" for custom p
weighted_metric_weights	number[]	null	Per-dimension weights (used when distance_metric = "weighted")
surface_mode	string	"interpolation"	Surface computation mode
cell	number	null	Cell size for interpolation grid (smaller → more vertices, higher accuracy)
interpolation_vertices	number	null	Number of interpolation vertices
zero_weight_fallback	string	"use_local_mean"	Zero-weight handling
boundary_policy	string	"extend"	Boundary handling policy
boundary_degree_fallback	boolean	null	Fall back to lower polynomial degree at boundaries when higher degrees fail
scaling_method	string	"mad"	Residual scaling method
auto_converge	number	null	Auto-convergence tolerance
missing	string	"error"	Policy for non-finite (NaN/Inf) values in input data
parallel	boolean	true	Enable parallel execution
outputs	string[]	[]	Optional fields: "diagnostics", "residuals", "weights", "gradient" (or "derivative"), "se", "sorted"
intervals	{ confidence?: number; prediction?: number }	disabled	Grouped confidence and prediction coverage levels.
cv	{ fractions: number[]; method?: string; k?: number }	disabled	Grouped candidate fractions, method, and folds; seed is an outer option
seed	number	unset	Non-negative safe-integer seed for reproducible CV folds, up to Number.MAX_SAFE_INTEGER.
retain_model	boolean	false	Retain training data, enabling result.predict()
custom_weights	number[]	null	Per-observation case weights — passed to fit(), not the options object

Options

fraction

fraction is the most important parameter: it controls the size of the local neighbourhood used at each point.

Range	Effect	Use case
0.1-0.3	Fine detail	Rapidly changing signals
0.3-0.5	Balanced	General purpose
0.5-0.7	Heavy smoothing	Noisy data
0.7-1.0	Very smooth	Trend extraction

iterations

iterations controls robustness to outliers, at the cost of speed.

Value	Effect	Performance
0	No robustness	Fastest
1-3	Moderate	Recommended
4-6	Strong	Contaminated data
7+	Very strong	Heavy outliers

weight_function

See: Weight Functions

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

robustness_method

See: Robustness

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

degree

See: Polynomial 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)

dimensions

See: Multivariate LOESS

Number of predictor dimensions. Set to match the number of columns in a multivariate x array.

• Any integer >= 1; 1 (default) is univariate

distance_metric

See: Multivariate LOESS

• "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")

weighted_metric_weights

See: Multivariate LOESS

Per-dimension weights, one per dimension declared in dimensions. Only used when distance_metric = "weighted"; setting distance_metric = "weighted" without providing this raises an error.

• null (default) — has no effect unless distance_metric = "weighted" is set
• A number[] of per-dimension weights, required when distance_metric = "weighted"

surface_mode

See: Polynomial Degree

Controls whether the local polynomial is evaluated at every query point or at a sparser grid of anchor vertices with Hermite cubic interpolation in between.

Mode	Behavior	Speed	Accuracy
"interpolation" (default)	Evaluate at vertices, interpolate between	Faster	Slight approximation
"direct"	Evaluate at every query point	Slower	Full precision

cell

Cell size for the interpolation grid, as a fraction of the data range. Smaller values place more vertices (denser grid), improving accuracy at the cost of speed. Only applies when surface_mode = "interpolation".

• null (default) — uses the library default (0.2)
• Any number in (0, 1]

interpolation_vertices

Caps the maximum number of interpolation vertices, overriding the count implied by cell. Only applies when surface_mode = "interpolation".

• null (default) — uses the library default (no explicit cap)
• Any integer >= 1

zero_weight_fallback

Behavior when all neighborhood weights are zero:

Option	Behavior
"use_local_mean" (default; aliases: "local_mean", "mean")	Use the mean of the neighborhood
"return_original" (alias: "original")	Return the original y value
"return_none" (alias: "none")	Return NaN

boundary_policy

See: Boundary Handling

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

boundary_degree_fallback

Whether to reduce the polynomial degree at boundary vertices when the requested degree can’t be fit there (e.g., not enough neighbours). Only applies when surface_mode = "interpolation".

• null (default) — uses the library default (enabled)
• true — falls back to a lower degree at boundaries
• false — raises an error instead of silently falling back

scaling_method

See: Scaling Methods

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

auto_converge

See: Robustness

Convergence tolerance for early stopping of robustness iterations. null (default) disables early stopping.

missing

Policy for handling non-finite (NaN/Inf) values in x/y (and custom weights):

Option	Behavior
"error" (default)	Throw an error if any value is non-finite
"drop"	Silently remove observations (rows) where any x dimension or y is non-finite before fitting

Note: A length mismatch between x and y always throws, even under "drop".

parallel

Enable multi-threaded execution via the Rayon-based web worker pool.

• true (default) — parallelizes the local regression fits
• false — forces single-threaded execution

outputs: se

See: Intervals

Computes hat-matrix statistics (effective degrees of freedom, leverage, delta1/delta2) in addition to standard errors.

outputs: diagnostics

See: Diagnostics

Include a Diagnostics object (RMSE, MAE, R², AIC/AICc, effective degrees of freedom) in the result. AIC/AICc/effective_df additionally require outputs: ["se"] (or confidence/prediction intervals) to be populated, since they depend on hat-matrix statistics.

outputs: residuals

Include per-point residuals (y - fitted) in the result.

outputs: weights

Include the final per-point robustness weights (from the last robustness iteration) in the result.

outputs: gradient

Each local polynomial fit (degree >= linear) already computes per-dimension coefficients internally, but only the fitted value is normally kept; this exposes that per-point gradient (rate of change of the smoothed surface, dimensions values per point, flattened) in result.gradient, enabling sensitivity/rate-of-change analysis at effectively no extra computation cost. Only supported when surface_mode is "direct" — the default "interpolation" mode only stores value+gradient at a sparse grid of vertices, not enough to reconstruct an exact per-point gradient, so fit() throws instead of silently leaving gradient as undefined. Omitted by default.

outputs: sorted

When selected in outputs, it reorders every result field (residuals, intervals, etc.) by x in an ascending manner, instead of in original input order. To get both orderings, sort the default result client-side (e.g. by the returned x array’s sort order) instead of calling fit() twice.

intervals.confidence

See: Intervals

Confidence level for the confidence interval around the mean response (e.g. 0.95). null (default) disables confidence intervals.

intervals.prediction

See: Intervals

Confidence level for the prediction interval for new observations (e.g. 0.95). null (default) disables prediction intervals.

CV Options

See: Cross-Validation

• cv.method: "kfold" (default) — fast, evaluates each candidate fraction over cv.k folds; "loocv" — slow, exhaustive leave-one-out cross-validation
• cv.k: Number of folds for k-fold CV. Ignored when cv_method = "loocv".
• cv.fractions: Candidate fractions to evaluate. Cross-validation is disabled unless this is set.
• seed: Non-negative safe-integer seed for reproducible k-fold shuffling, from 0 through Number.MAX_SAFE_INTEGER. null (default) uses a random seed; fractional, negative, non-finite, and unsafe values throw.

retain_model

See: Predict

Retains the fitted model’s training data, enabling result.predict(newX, options) to evaluate the fit at out-of-sample query points not in the training set. false (default) — no extra memory/copy cost unless requested.

custom_weights

See: Custom Weights

Per-observation weights, passed to fit() rather than the options object.

Result Structure

LoessResult

Field	Type	Description
x	Float64Array	x values (same order as input)
y	Float64Array	Smoothed y values
fraction_used	number	Fraction used (set or selected by CV)
iterations_used	number | undefined	Robustness iterations actually performed
standard_errors	Float64Array | undefined	Per-point SE (if "se" output)
confidence_lower	Float64Array | undefined	Lower confidence bounds
confidence_upper	Float64Array | undefined	Upper confidence bounds
prediction_lower	Float64Array | undefined	Lower prediction bounds
prediction_upper	Float64Array | undefined	Upper prediction bounds
residuals	Float64Array | undefined	Residuals (if "residuals" output)
robustness_weights	Float64Array | undefined	Robustness weights (if "weights" output)
cv_scores	Float64Array | undefined	CV score per tested fraction
diagnostics	Diagnostics | undefined	Fit metrics (if "diagnostics" output)
enp	number | undefined	Equivalent number of parameters (if "se" output)
trace_hat	number | undefined	Trace of hat matrix (if "se" output)
delta1	number | undefined	First delta statistic (if "se" output)
delta2	number | undefined	Second delta statistic (if "se" output)
residual_scale	number | undefined	Residual scale estimate (if "se" output)
leverage	Float64Array | undefined	Per-point hat-matrix diagonal (if "se" output)
gradient	Float64Array | undefined	Per-point local fit gradient, flattened (if "gradient" output, surface_mode = "direct" only)
dimensions	number	Number of predictor dimensions

Diagnostics

Field	Type	Description
rmse	number	Root Mean Squared Error
mae	number	Mean Absolute Error
r_squared	number	R-squared
residual_sd	number	Robust residual scale estimate (1.4826 * MAD)
effective_df	number | undefined	Effective degrees of freedom
aic	number | undefined	AIC
aicc	number | undefined	AICc

Predict

result.predict(newX, options) -> PredictOutput

Evaluates the fitted model at out-of-sample query points (flattened, dimensions values per point). Requires retain_model: true on the constructor before fit(), otherwise throws.

Example

const { Loess } = require('fastloess-wasm');

const x = new Float64Array([1, 2, 3, 4, 5]);
const y = new Float64Array([2.1, 4.0, 6.2, 8.0, 10.1]);

// Fit data
const model = new Loess({ fraction: 0.5 });
const result = model.fit(x, y);

console.log("Smoothed Y:", result.y);

Smoothed Y: Float64Array(5) [ 2.1, 4, 6.2, 8, 10.1 ]