fastlowess 0.99.4

High-performance LOWESS smoothing for Rust, Python, and R.

LOWESS Project

The fastest, most robust, and most feature-complete language-agnostic LOWESS (Locally Weighted Scatterplot Smoothing) implementation for Rust, Python, and R.

Overview

This monorepo contains a complete ecosystem for LOWESS smoothing:

• lowess - Core single-threaded implementation with no_std support
• fastLowess - Parallel CPU and GPU-accelerated wrapper with ndarray integration
• Python bindings - PyO3-based Python package
• R bindings - extendr-based R package

Documentation

[!NOTE]

📚 View the full documentation

Why this package?

Speed

The lowess project crushes the competition in terms of speed, wether in single-threaded or multi-threaded parallel execution.

Speedup relative to Python's statsmodels.lowess (higher is better):

Category	statsmodels	R (stats)	Serial	Parallel	GPU
Clustered	163ms	83×	203×	433×	32×
Constant Y	134ms	92×	212×	410×	18×
Delta (large–none)	105ms	2×	4×	6×	16×
Extreme Outliers	489ms	106×	201×	388×	29×
Financial (500–10K)	106ms	105×	252×	293×	12×
Fraction (0.05–0.67)	221ms	104×	228×	391×	22×
Genomic (1K–50K)	1833ms	7×	9×	20×	95×
High Noise	435ms	133×	134×	375×	32×
Iterations (0–10)	204ms	115×	224×	386×	18×
Scale (1K–50K)	1841ms	264×	487×	581×	98×
Scientific (500–10K)	167ms	109×	205×	314×	15×
Scale Large* (100K–2M)	—	—	1×	1.4×	0.3×

*Scale Large benchmarks are relative to Serial (statsmodels cannot handle these sizes)

The numbers are the average across a range of scenarios for each category (e.g., Delta from none, to small, medium, and large).

Robustness

This implementation is more robust than R's lowess and Python's statsmodels due to two key design choices:

MAD-Based Scale Estimation:

For robustness weight calculations, this crate uses Median Absolute Deviation (MAD) for scale estimation:

s = median(|r_i - median(r)|)

In contrast, statsmodels and R's lowess uses the median of absolute residuals (MAR):

s = median(|r_i|)

• MAD is a breakdown-point-optimal estimator—it remains valid even when up to 50% of data are outliers.
• The median-centering step removes asymmetric bias from residual distributions.
• MAD provides consistent outlier detection regardless of whether residuals are centered around zero.

Boundary Padding:

This crate applies a range of different boundary policies at dataset edges:

• Extend: Repeats edge values to maintain local neighborhood size.
• Reflect: Mirrors data symmetrically around boundaries.
• Zero: Pads with zeros (useful for signal processing).
• NoBoundary: Original Cleveland behavior

statsmodels and R's lowess do not apply boundary padding, which can lead to:

• Biased estimates near boundaries due to asymmetric local neighborhoods.
• Increased variance at the edges of the smoothed curve.

Features

A variety of features, supporting a range of use cases:

Feature	This package	statsmodels	R (stats)
Kernel	7 options	only Tricube	only Tricube
Robustness Weighting	3 options	only Huber	only Huber
Scale Estimation	2 options	only MAR	only MAR
Boundary Padding	4 options	no padding	no padding
Zero Weight Fallback	3 options	no	no
Auto Convergence	yes	no	no
Online Mode	yes	no	no
Streaming Mode	yes	no	no
Confidence Intervals	yes	no	no
Prediction Intervals	yes	no	no
Cross-Validation	2 options	no	no
Parallel Execution	yes	no	no
GPU Acceleration	yes*	no	no
no-std Support	yes	no	no

* GPU acceleration is currently in beta and may not be available on all platforms.

Validation

All implementations are numerical twins of R's lowess:

Aspect	Status	Details
Accuracy	✅ EXACT MATCH	Max diff < 1e-12 across all scenarios
Consistency	✅ PERFECT	Multiple scenarios pass with strict tolerance
Robustness	✅ VERIFIED	Robust smoothing matches R exactly

Installation

Currently available for R, Python, and Rust:

R (from R-universe, recommended):

install.packages("rfastlowess", repos = "https://thisisamirv.r-universe.dev")

Python (from PyPI):

pip install fastlowess

Or from conda-forge:

conda install -c conda-forge fastlowess

Rust (lowess, no_std compatible):

[dependencies]
lowess = "0.99"

Rust (fastLowess, parallel + GPU):

[dependencies]
fastLowess = { version = "0.99", features = ["cpu"] }

Quick Example

R:

library(rfastlowess)

x <- c(1, 2, 3, 4, 5)
y <- c(2.0, 4.1, 5.9, 8.2, 9.8)

result <- fastlowess(x, y, fraction = 0.5, iterations = 3)
print(result$y)

Python:

import fastlowess as fl
import numpy as np

x = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
y = np.array([2.0, 4.1, 5.9, 8.2, 9.8])

result = fl.smooth(x, y, fraction=0.5, iterations=3)
print(result["y"])

Rust:

use lowess::prelude::*;

let x = vec![1.0, 2.0, 3.0, 4.0, 5.0];
let y = vec![2.0, 4.1, 5.9, 8.2, 9.8];

let model = Lowess::new()
    .fraction(0.5)
    .iterations(3)
    .adapter(Batch)
    .build()?;

let result = model.fit(&x, &y)?;
println!("{}", result);

---

API Reference

R:

fastlowess(
    x, y,
    fraction = 0.5,
    iterations = 3L,
    delta = 0.01,
    weight_function = "tricube",
    robustness_method = "bisquare",
    zero_weight_fallback = "use_local_mean",
    boundary_policy = "extend",
    confidence_intervals = 0.95,
    prediction_intervals = 0.95,
    return_diagnostics = TRUE,
    return_residuals = TRUE,
    return_robustness_weights = TRUE,
    cv_fractions = c(0.3, 0.5, 0.7),
    cv_method = "kfold",
    cv_k = 5L,
    auto_converge = 1e-4,
    parallel = TRUE
)

Python:

fastlowess.smooth(
    x, y,
    fraction=0.5,
    iterations=3,
    delta=0.01,
    weight_function="tricube",
    robustness_method="bisquare",
    zero_weight_fallback="use_local_mean",
    boundary_policy="extend",
    confidence_intervals=0.95,
    prediction_intervals=0.95,
    return_diagnostics=True,
    return_residuals=True,
    return_robustness_weights=True,
    cv_fractions=[0.3, 0.5, 0.7],
    cv_method="kfold",
    cv_k=5,
    auto_converge=1e-4,
    parallel=True
)

Rust:

Lowess::new()
    .fraction(0.5)              // Smoothing span (0, 1]
    .iterations(3)              // Robustness iterations
    .delta(0.01)                // Interpolation threshold
    .weight_function(Tricube)   // Kernel selection
    .robustness_method(Bisquare)
    .zero_weight_fallback(UseLocalMean)
    .boundary_policy(Extend)
    .confidence_intervals(0.95)
    .prediction_intervals(0.95)
    .return_diagnostics()
    .return_residuals()
    .return_robustness_weights()
    .cross_validate(KFold(5, &[0.3, 0.5, 0.7]).seed(123))
    .auto_converge(1e-4)
    .adapter(Batch)             // or Streaming, Online
    .parallel(true)             // fastLowess only
    .backend(CPU)               // fastLowess only: CPU or GPU
    .build()?;

---

Result Structure

R:

result$x, result$y, result$standard_errors
result$confidence_lower, result$confidence_upper
result$prediction_lower, result$prediction_upper
result$residuals, result$robustness_weights
result$diagnostics, result$iterations_used
result$fraction_used, result$cv_scores

Python:

result.x, result.y, result.standard_errors
result.confidence_lower, result.confidence_upper
result.prediction_lower, result.prediction_upper
result.residuals, result.robustness_weights
result.diagnostics, result.iterations_used
result.fraction_used, result.cv_scores

Rust:

pub struct LowessResult<T> {
    pub x: Vec<T>,                           // Sorted x values
    pub y: Vec<T>,                           // Smoothed y values
    pub standard_errors: Option<Vec<T>>,
    pub confidence_lower: Option<Vec<T>>,
    pub confidence_upper: Option<Vec<T>>,
    pub prediction_lower: Option<Vec<T>>,
    pub prediction_upper: Option<Vec<T>>,
    pub residuals: Option<Vec<T>>,
    pub robustness_weights: Option<Vec<T>>,
    pub diagnostics: Option<Diagnostics<T>>,
    pub iterations_used: Option<usize>,
    pub fraction_used: T,
    pub cv_scores: Option<Vec<T>>,
}

---

LOESS vs. LOWESS

Feature	LOESS (This Crate)	LOWESS
Polynomial Degree	Linear, Quadratic, Cubic, Quartic	Linear (Degree 1)
Dimensions	Multivariate (n-D support)	Univariate (1-D only)
Flexibility	High (Distance metrics)	Standard
Complexity	Higher (Matrix inversion)	Lower (Weighted average/slope)

[!TIP] Note: For a LOESS implementation, use loess-project.

---

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md file for more information.

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)

at your option.

References

• Cleveland, W.S. (1979). "Robust Locally Weighted Regression and Smoothing Scatterplots". JASA.
• Cleveland, W.S. (1981). "LOWESS: A Program for Smoothing Scatterplots". The American Statistician.

Citation

If you use this software in your research, please cite it using the CITATION.cff file or the BibTeX entry below:

@software{lowess_project,
  author = {Valizadeh, Amir},
  title = {LOWESS Project: High-Performance Locally Weighted Scatterplot Smoothing},
  year = {2026},
  url = {https://github.com/thisisamirv/lowess-project},
  license = {MIT OR Apache-2.0}
}