AUXIL\
gpukernel_matrix.pro
top gpukernel_matrix
result = gpukernel_matrix(G1_gpu [, G2_gpu] [, KERNEL=integer] [, GMA=float] [, NSCALE=float] [, DEGREE=integer] [, BIAS=float])
Returns array of kernel functions for data matrices G1 and G2. G1 is an N x m array, G2 is an N x n array. Returned array is n x m. If G2 not present then returns a symmetric, positive definite kernel matrix
Parameters
- G1_gpu in required type={GPUHANDLE}
data matrix
- G2_gpu in optional type={GPUHANDLE}
data matrix
Keywords
- KERNEL in optional type=integer
the kernel used 0: linear 1: Gaussian (default) 2: polynomial 3: sigmoid
- GMA in out optional type=float
if not given and KERNEL=1, calculated from the data: GMA=1/(2*(nscale*scale)^2) where scale = average distance between observations in the input space, otherwise defaults to 1/N
- NSCALE in optional type=float
multiple of scale for GMA when KERNEL=1 (default 1.0)
- DEGREE in optional type=integer
degree of pylynomial kernel (default 2)
- BIAS in optional type=float
bias of polynomial or sigmoid kernel (default 1.0)
Author information
- Author:
Mort Canty (2009)
Other attributes
- Requires:
GPULIB
File attributes
Modification date: | Sun May 19 10:50:04 2013 |
Lines: | 94 |
Docformat: | rst rst |