function [xh,c] = SplineUChoose(y,k)
% SplineUChoose -- Estimate Optimal Bandwidth for Spline Kernel
% Usage
%   [xhat,risk] = SplineUChoose(y,k)
% Inputs
%   y       Noisy Data Normalized to Noise Level 1
%   k       Degree of Spline Penalty .
% Outputs
%   xhat	Estimate based on y with optimal lambda
%   risk    Estimate of Risk at this x
% Warning
%   must run script SplineInit before running this
% See Also:
%   SplineInit, SplineURisk
%

	global Spline_Pnty
	global Spline_Data
%
	f = fft(y);
	n = length(y);
	Spline_Data = abs(f/sqrt(n)).^2;
	Spline_Pnty = (2 .* sin(pi .* (0:(n-1)) ./ n)) .^ (2*k);
	Spline_Lmda = fminbnd('SplineURisk',0.,n.^2);
	xhh = f ./ (1 + Spline_Lmda.*Spline_Pnty);
	xh = real(ifft(xhh));
	%clf; subplot(211);
	%plot(1. ./ (1 + Spline_Lmda.*Spline_Pnty));
	%plot(xh);
	c  = SplineURisk(Spline_Lmda);

    
    

  %%  Part of Wavelab Version 850%  Built Tue Jan  3 13:20:41 EST 2006%  This is Copyrighted Material%  For Copying permissions see COPYING.m%  Comments? e-mail wavelab@stat.stanford.edu