[mlpack-svn] [MLPACK] #327: PCA Module Improvement
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Wed Feb 12 05:28:49 EST 2014
#327: PCA Module Improvement
---------------------------+------------------------------------------------
Reporter: siddharth.950 | Owner:
Type: defect | Status: new
Priority: major | Milestone:
Component: mlpack | Keywords:
Blocking: | Blocked By:
---------------------------+------------------------------------------------
The PCA module is currently not very efficient according to the
benchmarking system. I noticed one inefficiency in the implementation.
When the user specifies the new dimension for PCA, the module computes the
whole thing anyway, and then sheds off the remaining rows. Also, the
algorithm is currently using EigenValue Decomposition, which may not be
the optimal way to go about it according to this paper :
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.73.3994&rep=rep1&type=pdf.
This algorithm may make the implementation a lot faster that it already
is.
--
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/327>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed at Georgia Tech.
More information about the mlpack-svn
mailing list