Package Exports

ngraph.cw

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (ngraph.cw) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

ngraph.cw

Chinese Whispers graph clustering for ngraph. The algorithm is known for its speed (time-linear in the number of edges) and works best for large graphs.

I got algorithm converged for a graph with 200,000 nodes 500,000 edges after 8 iterations. Undirected graph with 2.2 million nodes, 4.5 million edges converged after 47 iterations.

You can play with small graph (254 edges, 77 nodes) here. It should converge after 2 iterations.

usage

var createWhisper = require('ngraph.cw');

// Graph is intance of ngraph.graph
var whisper = createWhisper(graph);

// The algorithm is iterative. We should continue running it until change
// rate within clusters is large:

var requiredChangeRate = 0; // 0 is complete convergence
while (whisper.getChangeRate() > requiredChangeRate) {
  whisper.step();
}
// When change rate is 0 (or very close to 0) we are done.
// Now we can query "class" of each node:
graph.forEachNode(printClass);

function printClass(node) {
  console.log(nodeId + ' belongs to ' + whisper.getClass(node.id));
}

By default ngraph.cw treats graph as undirected. You can change this via optional kind argument:

var createWhisper = require('ngraph.cw');

// Consider only neighbours with incoming edges:
var inDegreeWhisper = createWhisper(graph, 'in');

// Or outgoing edges:
var outDegreeWhisper = createWhisper(graph, 'out');

Iterating over clusters

So far we learned how to get class (or cluster) of every single node. ngraph.cw allows to do reverse operation too. This example shows how to iterate over each cluster and see nodes within:

var createWhisper = require('ngraph.cw');
var whisper = createWhisper(graph);
// assume we've done several iterations:
while (whisper.getChangeRate() > 0) {
  whisper.step();
}

// clusters is a Set object
var clusters = whisper.createClusterMap();

cluster.forEach(visitCluster)

function visitCluster(clusterNodes, clusterClass) {
  // each cluster has class identifier:
  assert(typeof clusterClass !== undefined);

  // And list of nodes within cluster:
  assert(Array.isArray(clusterNodes));
}

install

With npm do:

npm install ngraph.cw

license

MIT