sisl / BayesNets.jl / 172 / 2
85%
master: 85%

DEFAULT BRANCH: master
Ran
Files 20
Run time 2s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 94.525%. Remained the same
172.2

push

travis-ci

tawheeler 
Updated docs with an example of Gibbs sampling (#44)

* Definied the Gibbs sample and Variable elimination functions and added them to the package

* Wrote additional TODOs

* TODO comments

* Working on Gibbs sampling

* Working on Gibbs sampling...

* Added Notes to variable elimination, finished sample_posterior_finite

* Minor TODO comment on sampling_posterior_continuous

* Implemented selection the initial sample for burn in and getting the initial sample for the post burn in chain

* Implemented time_limit and sample_skip properly, added a temporary implementation for sample_posterior_continuous

* Added minor test cases for gibbs sampling, needs further work

* Fixed bugs in sampling and in tests, ran tests

* Added test for gibbs sampling from continuous distributions, fixed some bugs in gibbs sampling

* Added argument checking to gibbs sampling

* Wrote test for optional parameters and bad parameters for Gibbs sampling.  Fixed related bugs in Gibbs sampling

* Added some testing for discrete bayes nets, a folder that will be used for comparisons (will be removed before pull request) and some comments in sampling.jl

* Implemented plotting and comparisons in gibbs_sampling_comparison.jl, fixed a few bugs in gibbs sampling in sampling.jl

* Minor parameter tweaking on test cases

* Added samples on x axis plot to comparison code

* Added code to profile the sampler easily

* Added comparison from the book for Gibbs versus likelihood weighted and beginning to work on caching for Gibbs sampler

* Added caching to gibbs sampler (store finite distributions and markov blankets

* Added test case for continuous distributions

* Implemented Metropolis-Hastings for sampling continuous distributions, not sure about the proposal or the number of steps to run the chain

* Comments on TODOs

* Commented out something I was using last commit, undoing that

* Comments and removed unnecessary operation in MH

* sample_posterior renamed to sample_posterior! and the function modifies instead of returns, also renaming of variables in sample_posterior_continuous

* Removed variable elimination stuff because it will probably not be gotten to

* Moved all the gibbs sampling stuff to its own file

* Moved Gibbs unit tests to their own file

* Bug fix in code that produces graphs for gibbs sampling in the continuous case

* Added markov_blanket_cpds to bayes_nets.jl

* Added code for testing and making plots for Metropolis Hastings, this is a checkpoint for debugging it

* Working on MH plotting code, this is a checkpoint

* Fixed bug in MH_comparison.jl

* Finished MH_comparison.jl

* Added another test case to gibbs_sampling_comparison.jl

* Added a decent description of the gibbs sampler

* Added new entry point for Gibbs sampling.  Base.rand(bn::BayesNet, config::GibbsSampler, nsamples::Integer)

* Implementing IA2RMS, this is a checkpoint

* Implementing IA2RMS, checkpoint 2

* Beginning to debug IA2RMS

* Debugging IA2RMS checkpoint

* Made an abstract type BayesSampler, also improved the efficiency of the caching and added a markov blanket function

* Modified to take an average over many runs

* Adding documentation and making stylistic changes

* Clearning up before pull request

* Bug fix when making style changes

* Now thrown an error if the pdf of the inital sample supplied is zero

* Adding support for julia v0.4

* Removed deprecation warnings in julia 0.4

* Fixing deprecation errors in julia 0.5

* Added an example of gibbs sampling to the doc, also corrected an error in a message to the user (sample_skip was renamed to thinning but this was not updated)

846 of 895 relevant lines covered (94.53%)

5386.12 hits per line

Source Files on job 172.2