Giter Site home page Giter Site logo

bentobox's Introduction

Bento Box MCMC

Authors

  • Dan Foreman-Mackey (NYU)
  • Jonathan Goodman (NYU)
  • David W. Hogg (NYU)
  • Fengji Hou (NYU)

License

This is an academic project; the software doesn't even work. Use at your own risk. Copyright 2013 the Authors.

Abstract

Most MCMC methods perform very badly when the density function being sampled is multi-modal, especially when the modes are separated by large regions of small or vanishing density; this is generic in many data-analysis problems of interest to scientists. Here we propose a sampler that solves this problem by recursively splitting the parameter space until each sub-space contains at most one mode, or at least some easy-to-sample distribution. The modes are discovered by initializing standard samplers at prior samplings; we use the housekeeping data from an ensemble sampler to provide mode (or cluster) label information. Each parameter sub-space split is performed by a support vector machine. The (now easy-to-make) sub-space samplings are recombined into a complete sampling by importance sampling on the total posterior integral within each sub-space; this requires computation of the marginalized likelihood (or Bayes factor) in every sub-space; we propose a simple method for this that works well for uni-modal density functions (which we have, by construction). In its naivest form, the method requires proper priors, and is faster if the prior PDF can be sampled quickly or simply; again this is common for many data-analysis problems of interest in scientific applications. We demonstrate the value of the method on a data-analysis problem in exoplanet astronomy. It outperforms [which other methods] by [how much].

bentobox's People

Contributors

davidwhogg avatar

Watchers

 avatar  avatar  avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.