High-Dimensional Space Projections

TypeScript library for sparse and dense high-dimensional space projections.

Usage

TypeDocs

https://ibm.github.io/projections/

Examples

Multi-Dimensional Scaling (PivotMDS)

import * as hdsp from "hdsp";
hdsp.PivotMDS.project(
    [
        [1,0,0,1,0,0,0,1],  // feature vector 1
        [0,1,1,0,1,1,1,1],  // feature vector 2
                            // ...
        [0,0,0,1,0,1,0,1]   // feature vector n
    ],
    4, // number of pivots
    2  // output dimensionality
);

Multi-dimensional Scaling by Stochastic Gradient Descent (MDSSGD)

import * as hdsp from "hdsp";
hdsp.MDSSGD.project(
    [
        [1,0,0,1,0,0,0,1],  // feature vector 1
        [0,1,1,0,1,1,1,1],  // feature vector 2
                            // ...
        [0,0,0,1,0,1,0,1]   // feature vector n
    ],
    2  // output dimensionality
);

Sparse Multi-dimensional Scaling by Stochastic Gradient Descent (Sparse-MDSSGD)

import * as hdsp from "hdsp";
hdsp.SparseMDSSGD.project(
    5, 4,                   // n, m 
    [0, 1, 2, 3],           // edges: from
    [1, 2, 3, 4],           // edges: to
    [1, 1, 1, 1],           // edges: values
    2,                      // output dimensionality
    4,                      // number of pivots
    1,                      // seed
    100,                    // max iterations
    0.001                   // minimum layout change to continue
);

Procrustes Analysis (2D)

import * as hdsp from "hdsp";
hdsp.Procrustes.transform(
    [[0,0],[0,-2],[-1,-1]],  // input
    [[0,0],[0,2],[1,1]]      // reference
);

Developers

Code repository

https://github.com/IBM/projections

Build / Release Instructions

1. tsc
2. typedoc --out docs
3. touch docs/.nojekyll
4. npm version patch
5. npm publish
6. git push