TADM has been updated for the most recent versions of PETSc (2.3.3) and TAO (1.9). This should actually make it considerably easier to install TADM because the new PETSc installation is considerably smoother than for past versions.
The Python front end has been extended to work with ranking as well as classification, and there are example data sets for use with the Python frontend to be found in tadm/data. See the README and docs/install.txt.... read more
A bug was introduced in the last version which meant that the classification models produced using the Python interface produced incorrect results. (It was a Python indentation error... argh.) Version 0.9.7b fixes this bug.
Models built with the low level numerical format for event files are unaffected by this bug.
I've made some minor updates to enhance the Python interface to TADM, specifically to allow one to create text-based feature files for ranking tasks (so one doesn't need to manage feature mapping).
Also, people might find the following homework from my compling 2 class useful for learning how to use TADM:
The first version of the code with the TADM name and LGPL license is now available for download. It is otherwise essential ly the same as Rob Malouf's original code:
TADM has moved to Sourceforge to facilitate collaboration on the development of the package. Initially, this collaboration consists of Rob Malouf (San Diego State University), Jason Baldridge (University of Texas at Austin), and Miles Osborne (University of Edinburgh). Please let us know if you are also interested in working on either the code or documentation.... read more