Quick Start

Get up and running with LOESS in minutes.

Basic Smoothing

Smooth a noisy sine wave — the kind of signal where LOESS shines. Each example recovers the underlying trend from 100 points of Gaussian noise.

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

// 100-point noisy sine wave
const n = 100;
const x = Float64Array.from({ length: n }, (_, i) => i * 2 * Math.PI / (n - 1));
const y = Float64Array.from(x, (xi, i) => Math.sin(xi) + (((i * 7 + 3) % 17) / 17 - 0.5) * 0.6);

const model = new Loess({ fraction: 0.3, iterations: 3 });
const result = model.fit(x, y);

console.log(`First smoothed: ${result.y[0].toFixed(4)}  (true: ${Math.sin(x[0]).toFixed(4)})`);

First smoothed: 0.0281  (true: 0.0000)

---

With Confidence Intervals

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

const n = 100;
const x = Float64Array.from({ length: n }, (_, i) => i * 2 * Math.PI / (n - 1));
const y = Float64Array.from(x, (xi, i) => Math.sin(xi) + (((i*7+3)%17)/17-0.5)*0.6);

const model = new Loess({
    fraction: 0.5,
    iterations: 3,
    outputs: ["diagnostics"],
    intervals: { confidence : 0.95, prediction : 0.95 }
});
const result = model.fit(x, y);

console.log("Smoothed (first 5):", [...result.y.slice(0, 5)].map(v => v.toFixed(4)));
console.log("CI lower (first 5):", [...result.confidence_lower.slice(0, 5)].map(v => v.toFixed(4)));
console.log("CI upper (first 5):", [...result.confidence_upper.slice(0, 5)].map(v => v.toFixed(4)));
console.log("R2:", result.diagnostics.r_squared.toFixed(4));

Smoothed (first 5): [ '0.1118', '0.1389', '0.1702', '0.2064', '0.2445' ]
CI lower (first 5): [ '-0.0075', '0.0204', '0.0524', '0.0894', '0.1284' ]
CI upper (first 5): [ '0.2311', '0.2575', '0.2879', '0.3233', '0.3605' ]
R2: 0.8860

---

Handling Outliers

LOESS can robustly handle outliers through iterative reweighting:

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

const xOut = new Float64Array([1, 2, 3, 4, 5, 6]);
const yWithOutlier = new Float64Array([2.0, 4.0, 6.0, 50.0, 10.0, 12.0]);

const model = new Loess({
    fraction: 0.7,
    iterations: 5,
    robustness_method: "bisquare",
    outputs: ["weights"]
});
const result = model.fit(xOut, yWithOutlier);

// Outliers will have low robustness weights
result.robustness_weights.forEach((w, i) => {
    if (w < 0.5) {
        console.log(`Point ${i} is likely an outlier (weight: ${w.toFixed(3)})`);
    }
});

Point 3 is likely an outlier (weight: 0.000)

---

Streaming Mode

For datasets too large to fit in memory, stream them in fixed-size chunks with overlap.

const { StreamingLoess } = require('fastloess');

const n = 5000;
const x = Float64Array.from({ length: n }, (_, i) => i * 10 * Math.PI / (n - 1));
const y = Float64Array.from(x, (xi, i) =>
    Math.sin(xi / Math.PI) * Math.exp(-xi / 30) +
    (((i * 7 + 3) % 17) / 17 - 0.5) * 0.3
);

const model = new StreamingLoess(
    { fraction: 0.2 },
    { chunk_size: 1000, overlap: 100, merge_strategy: 'weighted_average' }
);

const chunk_size = 1000;
for (let start = 0; start <= 4000; start += chunk_size) {
    const end = Math.min(start + chunk_size, n);
    model.process_chunk(x.slice(start, end), y.slice(start, end));
}
const result = model.finalize();
console.log(`Smoothed ${result.y.length} points in streaming mode`);

Smoothed 100 points in streaming mode

---

Next Steps

Topic	Link
How LOESS works	Concepts
All parameters explained	API Reference
Batch vs Streaming vs Online	Execution Modes
Polynomial degree choices	Degree
Multivariate smoothing	Dimensions
Edge handling	Boundary
Outlier handling in depth	Robustness
Full API per language	API Reference