https://github.com/dfm/emcee.js

Just a slick little Markov chain Monte Carlo sampler written in Javascript...
https://github.com/dfm/emcee.js

Last synced: about 1 year ago
JSON representation

Just a slick little Markov chain Monte Carlo sampler written in Javascript...

• Host: GitHub
• URL: https://github.com/dfm/emcee.js
• Owner: dfm
• License: mit
• Created: 2012-07-06T17:04:00.000Z (almost 14 years ago)
• Default Branch: main
• Last Pushed: 2020-06-12T18:17:42.000Z (almost 6 years ago)
• Last Synced: 2025-03-27T10:21:38.836Z (about 1 year ago)
• Language: JavaScript
• Homepage: http://danfm.ca/emcee.js/
• Size: 84 KB
• Stars: 7
• Watchers: 4
• Forks: 1
• Open Issues: 3
• Metadata Files:
  • Readme: README.markdown
  • License: LICENSE.markdown

Awesome Lists containing this project

README

          
emcee.js

========

This a Javascript implementation of [emcee](http://danfm.ca/emcee).

Usage

-----

1. Write down the density that you want to sample:

         var lnprobfn = function (x) {

             var i, result = 0.0;

             for (i = 0; i < x.length; i++) result -= 0.5 * x[i] * x[i];

             return result;

         };

2. Initialize an `EnsembleSampler` object:

         var sampler = new emcee.EnsembleSampler(lnprobfn);

3. Make an initial guess at positions for 100 walkers in 3 dimensions,

   in this particular case, it's going to be a small Gaussian ball:

         var initialPosition = emcee.smallBall(100, [0.1, 2, -0.5], [1.0, 0.1, 0.5]);

4. Run the heck out of it and take 1000 samples:

         sampler.runMCMC(initialPosition, 1000);

5. The acceptance fraction is `sampler.acceptanceFraction` and the chain

   is stored in `sampler.chain`.

6. Now make some sort of **sick front end**!

Example

-------

To see an example, run `python -m SimpleHTTPServer 8000` in the root directory of this

repository and navigate to [localhost:8000/examples](http://localhost:8000/examples).