%PLOTF Plot feature distribution, special version
% 
%   h = PLOTF(A,N)
% 
% Produces 1-D density plots for all the features in dataset A. The 
% densities are estimated using PARZENML. N is the number of 
% feature density plots on a row. 
% 
% See also DATASETS, PARZENML

% Copyright: R.P.W. Duin, duin@ph.tn.tudelft.nl
% Faculty of Applied Sciences, Delft University of Technology
% P.O. Box 5046, 2600 GA Delft, The Netherlands

% $Id: plotf.m,v 1.6 2009/11/13 08:54:18 davidt Exp $

function h_out = plotf(a,n,z)
  %DXD make a standard setting for n, I'm getting crazy!
  if nargin<2
    n = 1;
  end

		
  if ~isdataset(a)
    a = prdataset(a,1); % solves a lot of problems
  end
  [m,k,c] = getsize(a);

	% Define the color for each of the classes:
  if c == 1
    clrmap = [0 0 1];
	elseif c == 2
		clrmap = [0 0 1; 1 0 0];
	else
		clrmap = hsv(c);
	end

	% Make subplots for each feature, so a grid of p x q subplots is
	% defined
	h = [];
	if k >= n
		p = ceil(k/n); q = n;
	else
		p = k; q = 1;
  end

  if isempty(getfeatlab(a))
    a = setfeatlab(a,[1:k]');
  end
  % Get the feature names
  feats = getfeatlab(a,'string');
	%DXD what happens here?!
  %RD If feature labels are scalars to single characters it might be
  %nicer to put 'Feature ' in front of it.
	if size(feats,2) == 1
		feats = [repmat('Feature ',size(feats,1),1) feats];
	end
	if isempty(feats)
		feats = num2str((1:k)');
	end

	% Make the plot for each of the features:
	for j = 1:k
		b = a(:,j);
		s = zeros(1,c);
		d = zeros(121,c);
		bb = [-0.10:0.01:1.10]' * (max(b)-min(b)) + min(b);
		ex = 0;
		% Make a density estimate of each of the classes:
		for i = 1:c
			I = findnlab(a,i);
			D = +distm(bb,b(I,:));
      if nargin < 3
        s(i) = parzenml(b(I,:));
      else
        s(i) = z(i);
      end
			% Compute the density function
			d(:,i) = sum(exp(-D/(s(i).^2)),2)./(length(I)*s(i));;
		end
		% Create the subplots with the correct sizes:
    if p==1 && q==1
      % avoid subplots in case of single plot
      plot(bb,zeros(size(bb)),'w.');
      hold on;
    else
      subplot(p,q,j)
      plot(bb,zeros(size(bb)),'w.');
      hold on
    end
		h = [];
		% Scatter the data and plot the density functions for each of the
		% classes:
		for i = 1:c
			I = findnlab(a,i);
			hh = plot(b(I),zeros(size(b(I))),'x',bb,+d(:,i));
			set(hh,'color',clrmap(i,:));
			h = [h;hh];
		end
		legend(h(1:2:end)',num2str(getlablist(a))); %does not work properly
		title([getname(a) ': ' feats(j,:)]);
		V = axis;
		axis([bb(1) bb(end) V(3) V(4)]);
		set(gca,'xtick',[]);
		set(gca,'ytick',[]);
		xlabel(feats(j,:));
		hold off
	end

	% The last details to take care of:
	if k == 1, title(''); end
	if nargout > 0
		h_out = h;
	end

	return