GuidelinesForContributors

1. summary Best practices to follow when contributing data to pmtkdata

The PMTK3 loadData function automatically downloads required data sets from the pmtkData website. To ensure that this works for new data sets, please follow these guidelines:

  *  Each data set must live in in its own folder and have exactly one .mat file. Store any processing scripts and source data in this folder as well. 

  * The name of the .mat file and the folder must be the same. 

  * Each data set must have a foo-meta.m file stored in the [http://pmtkdata.googlecode.com/svn/trunk/meta/ pmtkData/meta] directory, where "foo" is the name of the data set. This file stores "meta" information used to generate [http://pmtkdata.googlecode.com/svn/trunk/docs/dataTable.html this table].

  * Each line of the foo-meta.m file begins with a % followed by a space, and contains a [http://code.google.com/p/pmtk3/wiki/PMTKtags PMTK tag] along with its data. [http://pmtkdata.googlecode.com/svn/trunk/meta/20news_w1000-meta.m Here] is an example. 

  * Check the folder in to svn. 

  * Zip up the new folder by running the following PMTK command:

  The [http://code.google.com/p/pmtk3/source/browse/trunk/localUtil/supportTools/refreshZipFiles.m refreshZipFiles] function automatically excludes .svn directories before creating the zip file, which is stored inside the foo directory. If you are zipping up a folder manually, make sure not to include the svn info, either by doing this before you check the folder in, or by running {{{ svn export }}}.

 * Check the folder in to svn again, to include the .zip file.

  * Test auto-downloding with {{{loadData('foo') }}} making sure the folder you just created is not on your MATLAB path. 

  * Run {{{generatePmtkDataTable}}} to update the [http://pmtkdata.googlecode.com/svn/trunk/docs/dataTable.html list of datasets].

  * That's it