[mlpack-svn] [MLPACK] #167: Different solution for HRectBound::MinDistance(..., std::vector<>)
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Wed Nov 30 11:35:34 EST 2011
#167: Different solution for HRectBound::MinDistance(..., std::vector<>)
-------------------------------------------------+--------------------------
Reporter: rcurtin | Owner: nslagle
Type: defect | Status: new
Priority: major | Milestone: MLPACK 1.0
Component: MLPACK | Resolution:
Keywords: bound tree policy template methods | Blocking:
Blocked By: 136 |
-------------------------------------------------+--------------------------
Changes (by nslagle):
* blockedby: => 136
Comment:
This was an issue I had intended to discuss during Monday's meeting, but,
alas, I forgot.
I can imagine needing a filtering on a general bound in any setting
requiring some estimate conditioned on a proper subset of indices. I
chose not to implement this feature in a separate class because of the
increased memory footprint we could encounter. That is, suppose the user
intends to measure data containing hundreds of thousands or millions of
dimensions. To condition with an additional bounds class, the memory
footprint grows considerably. Consider also if the user intends to try
myriad conditioning subsets, leaving all bound objects in memory. Yes,
the user probably is making a mistake, but herein is another opportunity
to offer a little protection to the user.
This actually touches the more general issue I raised in ticket:136; it
would make sense to do whatever is possible to reduce the memory footprint
in many places in our code by reusing objects as much as possible.
Also, this is somewhat in the spirit of ticket:104, in that we need
greater configurability of existing objects.
--
Ticket URL: <https://trac.research.cc.gatech.edu/fastlab/ticket/167#comment:1>
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