RBFPREIMG2 |
RBF pre-image problem by Gradient optimization.
Synopsis:
z = rbfpreimg2(model)
z = rbfpreimg2(model,options)
Description:
z = rbfpreimg2(model) it uses gradient method to solve
the pre-image problem for the Radial Basis Function (RBF)
kernel. The function 'fminunc' of the Matlab Optimization
toolbox is exploited for 1D search along the gradient
direction.
z = rbfpreimg2(model,options) use to specify the control
parameters of the gradient optimization.
Input:
model [struct] Kernel expansion:
.Alpha [num_data x 1] Weight vector.
.sv.X [dim x num_data] Vectors determining the kernel expansion.
.options.arg [1x1] Argument of the RBF kernel (see 'help kernel').
options [struct] Control parameters:
.min_improvement [1x1] Minimal allowed improvement of the objective
function in two consecutive steps (default 1e-3).
options.start_t [1x1] Starting value of the 1D search procedure
(default 1e-3).
Output:
z [dim x 1] Found preimage.
See also
RBFPREIMG, RBFPREIMG3, RSRBF, KPCAREC.
About: Statistical Pattern Recognition Toolbox
(C) 1999-2003, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
10-jun-2004, VF
03-dec-2003, VF