rnx

function rnx(cr, k): number;

Defined in: metrics/neighborhood.ts:241

Neighbourhood preservation rescaled so chance is 0 and perfect is 1.

Preferred over qnx when comparing across different k or dataset sizes, since those cancel out.

• Needs: high-dimensional data and projection. No labels.
• Range: [0, 1] in practice, higher is better; 0 is a random projection.
• Cost: O(1), from an O(N² log N) pass.

Parameters

Parameter	Type
cr	CoRanking
k	number

Returns

number

See

Lee & Verleysen, Neurocomputing 72 (2009) https://doi.org/10.1016/j.neucom.2008.12.017

Example

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

const a = analyze(data, projection);   // number[][] straight in
rnx(a.coRanking, 20);  // 0.5172 on a PCA projection
// the same call on a random projection of the same points: 0.0042