graphslam is an optimization-based SLAM algorithm. It describes the SLAM problem using a graph, which is type of bayesian network. More details of graphslam can be found in this paper and this course.
This matlab version is a modification from Giorgio Grisetti's Octave code. The original Octave version can be found [here] (http://www.dis.uniroma1.it/~grisetti/teaching/lectures-ls-slam-master/web/lectures/practicals/ls-slam.tgz)
All the license information is included in lsslam folder.