residualVariance

function residualVariance(m): MetricResult;

Defined in: metrics/correlation.ts:38

The share of distance variance the projection fails to explain.

Simply 1 - r² of pearsonR, restated so that lower is better and the number reads as “proportion of structure lost”.

• Needs: high-dimensional data and projection. No labels.
• Range: [0, 1], lower is better.
• Cost: O(1), from an O(N²·D) pass.

Parameters

Parameter	Type
m	DistanceMoments

Returns

MetricResult

See

Tenenbaum, de Silva & Langford, Science 290 (2000) https://doi.org/10.1126/science.290.5500.2319

Example

import { analyze, residualVariance } from "@saehrimnir/sickle";

const a = analyze(data, projection);
residualVariance(a.moments).value;  // 0.0066 — 1 - r², so lower is better