Sammon

Sammon Mapping is a non-linear approach for mapping higher-dimensional space to a space of lower dimensionality, which attempts to preserve structure.

How It Works

Sammon Mapping is a variation of metric MDS that uses a specific cost function (Sammon's stress) which heavily penalizes errors in preserving smaller distances over larger ones.

Why or When to Use

Use Sammon Mapping when preserving local distances (small distances between nearby points) is more important than global distances, often resulting in better cluster separation.

Example

How-to (Code)

javascript

import * as druid from "@saehrimnir/druidjs";

const data = [
  /* ... multi-dimensional data ... */
];

// 1. Initialize the algorithm
const sammon = new druid.SAMMON(data);

// 2. Compute the projection
const projection = sammon.transform();

Reproducibility

Passing a seed pins the result for a given engine and library build, but not across environments: the gradient step is chaotic, so a difference in the last bit — between two JavaScript engines, or between the WASM kernel and its JS fallback — grows into a visibly different layout within a few dozen iterations. ECMAScript specifies Math.pow and Math.exp as implementation-approximated, so such differences cannot be ruled out.

The effect is equivalent to changing the seed rather than a loss of quality; see UMAP for the measured detail. MDS and SMACOF are stable alternatives when a layout has to reproduce everywhere.

Sammon ​

How It Works ​

Why or When to Use ​

Example ​