[mlpack-svn] [MLPACK] #288: run GMM::Estimate with weighted data
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Thu May 16 08:42:14 EDT 2013
#288: run GMM::Estimate with weighted data
------------------------+---------------------------------------------------
Reporter: hanslovsky | Owner:
Type: wishlist | Status: new
Priority: major | Milestone:
Component: mlpack | Keywords: GMM Estimate weighted
Blocking: | Blocked By:
------------------------+---------------------------------------------------
Having a GMM::Estimate that accepts weighted data would be a great
enhancement of the GMM class. To make my wish more clear, I want to use
GMM on probability maps, each voxel value in the range [0, 1]. So far we
are doing clustering on a thresholded binary map (including all voxels
with a value above a certain threshold). The data for the GMM is the
coordinates of the according voxels.
One way of using the probability map instead of the binary map would be
replicating the number of data points according to the probability.
However, that would blow up the amount of data by a lot. A GMM::Estimate
that takes the data (coordinates) and weights (probabilities) as input
would be the best solution.
If a solution to this already exists, please point it out to me.
Otherwise, if a GMM with weighted data is possible, having it in mlpack
would mean a great improvement for me.
--
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/288>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed by the FASTLAB at Georgia Tech under Dr. Alex Gray.
More information about the mlpack-svn
mailing list