================================================================
= =
= Functions for Mean-Shift clustering of top-t rankings =
= =
================================================================
Available at: www.stat.washington.edu/mmp/intransitive.html
First version of this package written by Le Bao (2008)
MS.m re-written by Marina Meila MMP@stat.washington.edu (2012)
Citation:
An Exponential Model for Infinite Rankings
Marina Meila, Le Bao; Journal of Machine Learning Research 11(Dec):3481−3518, 2010.
Please cite the work above if you use our code.
@Article{MBao:jmlr-infinite10,
author = {Marina Meil\u{a} and Le Bao},
title = {An exponential family model over infinite rankings},
journal = {Journal of Machine Learning Research},
year = {2010},
volume = {10},
pages = {3481--3518},
month = {December},
url = {http://jmlr.csail.mit.edu/papers/v11/meila10a/meila10a.pdf},
}
For more details see also:
@TechReport{BaoM:computing-exam-tr08,
author = {Le Bao and Marina Meil\u{a}},
title = {Clustering permutations by Exponential Blurring Mean-Shift},
institution = {University of Washington},
year = {2007},
type = {Department of Statistics},
number = {524},
address = {},
month = {},
note = {},
annote = {}
}
===============================================================
Code overview
Type >help at the Matlab prompt to see a detailed
description of each function
Scripts
-------
main.m -- runs a clustering experiment comparing various
algorithms
test_MS.m -- compares the running times of MS.m and MS.0 (MMP)
Core functions
--------------
MS.m -- mean-shift algorithm by MMP
MS0.m -- original mean-shift code by Le Bao
MS_infinite.m -- mean-shift algorithm for infinite permutations by
Le Bao
MBC_infinite.m -- model-based (EM) clustering
K_means.m -- K-means clustering
Helper functions
----------------
clustering_error.m
compute_confusion.m
dKdiff.m -- Kendall distance, calls dK.m
dK.m -- Kendall distance
make_Qp.m
sample_from_theta.m
--------------------------------------------------------------------
README for First version written by Le Bao (2008)
To run the code:
1. change working directory to the folder containing the codes
2. open main.m, choose model_option and simulation design
The core functions are MS_infinite() for mean-shit clustering,
MBC_infinite() for model based clustering, and kmeans().