mljs / kmeans Goto Github PK

I tried to use ml-kmeans in a TypeScript project, but it appears that kmeans I import is not a function, generating .d.ts file may help.

Error message that states, "K should be a positive integer bigger than the number of points" should say "smaller", rather than "bigger". Line 76 src/kmeans.js

Happy to fix this and submit a pull request, if preferred. Didn't think an error message typo may be worthy of a PR. :-)

function kmeans(data, k, options)

options.initialisation

• array of your current centers
• 'random' (default value) : we take randomly k points of the data

Hi,

I'm using Rollup to build an es6 module and when I go to build my files, I get the following error:

[!] Error: 'squared' is not exported by node_modules/ml-kmeans/node_modules/ml-distance-euclidean/euclidean.js
npm ERR! code ELIFECYCLEp/rollup/wiki/Troubleshooting#name-is-not-exported-by-module
npm ERR! errno 1kmeans/src/initialization.js (2:9)

I realize this could probably be handled using rollup-plugin-commonjs, but it looks like the issue has already been partially addressed in version 4.1.

Your distance-euclidean repo builds a directory lib-es6 but this is not present in node_modules/ml-kmeans/node_modules/ml-distance-euclidean/ so it ends up using the default commonjs version.

Am I missing something here? Has ml-kmeans been tested with an es6 build?

Thanks,
Josh

• weighted kmeans
• triangle inequality
• kernel kmeans
• kmedian

• kmeans++
• Random partition

First of all, thanks for the amazing library.

1. How cumbersome would it be to add the wcss to the KMeansResult?
2. Would it be possible to implement a predict method?

Thanks again!

In the userguide it seems it works on with x,y points. Is this correct ?

I would like to kluster r,g,b points.

I'd like to refactor the library to a class, with an API similar to sklearn: http://scikit-learn.org/stable/modules/generated/sklearn.cluster.KMeans.html

I'll migrate to TypeScript at the same time.

@mljs/collaborators WDYT?

• Convert code to typescript
• Generate tsdoc

Sometimes KMeansResult's centroids value is NaN for both x,y (Two-dimensional array) and for this centroid error value is NaN.