aucLogRnx

function aucLogRnx(cr): number;

Defined in: metrics/neighborhood.ts:270

One number summarising neighbourhood preservation at every scale.

Averages rnx over all k on a logarithmic scale, so small neighbourhoods count for more than large ones. Use it when you do not want to commit to a k.

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

Parameters

Parameter	Type
cr	CoRanking

Returns

number

See

Lee et al., Neurocomputing 169 (2015) https://doi.org/10.1016/j.neucom.2014.12.095

Example

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

const a = analyze(data, projection);   // number[][] straight in
aucLogRnx(a.coRanking);  // 0.4658 — no k to choose