[mlpack-svn] [MLPACK] #136: remove blocking system calls from computationally intensive inner loops

[MLPACK Trac]

#136: remove blocking system calls from computationally intensive inner loops
--------------------------+-------------------------------------------------
  Reporter:  nslagle      |        Owner:     
      Type:  enhancement  |       Status:  new
  Priority:  major        |    Milestone:     
 Component:  MLPACK       |   Resolution:     
  Keywords:               |     Blocking:  167
Blocked By:               |  
--------------------------+-------------------------------------------------

Comment (by rcurtin):

 No, we can't reduce kernels to be singletons.  With the GaussianKernel,
 you can change the bandwidth.  If you really, really, really need a static
 Evaluate() for some reason, then write template<int bandwidth>
 GaussianKernel(), which, of course, can't have a double bandwidth because
 template parameters can be ints, bools, or typenames.

 I'm not seeing the exact usage of this you're referring to, anyway.  If
 you're running, say, Kernel PCA, then you only need one instantiation of
 the kernel (held inside the `KernelPCA` class).  You can modify it as
 needed, but, there is only one of them.

-- 
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/136#comment:3>
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.