[mlpack-svn] [MLPACK] #273: Support for instantiated kernels and metrics is somewhat poor
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Thu Apr 25 16:52:04 EDT 2013
#273: Support for instantiated kernels and metrics is somewhat poor
---------------------+------------------------------------------------------
Reporter: rcurtin | Owner:
Type: defect | Status: new
Priority: major | Milestone: mlpack 1.1.0
Component: mlpack | Keywords: metric, kernel, instantiated, mahalanobis distance
Blocking: | Blocked By:
---------------------+------------------------------------------------------
In many situations it is useful to have a kernel or metric that saves
state. One instance is the Mahalanobis distance, which must store the
covariance matrix. Another instance is the polynomial kernel and
hyperbolic tangent kernel, which each have runtime parameters that cannot
be template parameters (doubles not ints).
This means that all methods that depend on some sort of kernel or metric
must be able to seamlessly work with instantiated kernels and metrics.
Right now, all things work well with kernels and metrics that do not store
any state (that is, they can be called statically).
This is something of a master ticket aimed at making this support correct
and consistent across mlpack. There will probably be some reverse
compatibility issues so the non-reverse-compatible changes should be saved
until mlpack 1.1.0.
--
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/273>
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