1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234
|
function pagerankdemo (steps)
% PAGERANKDEMO draw a 6-node web and compute its pagerank
%
% PAGERANKDEMO draws the 6-node "tiny web" in Section 2.11 of "Numerical
% Computing with MATLAB", by Cleve Moler, SIAM, 2004. It then simulates the
% computation of Google's PageRank algorithm, by randomly selecting links to
% traverse. If a link is traversed, the edge and the target node are displayed
% in red. If the "random surfer" jumps to an arbitrary page, the target node
% is displayed in blue. The number of hits at each node, and the page rank
% (in %) are displayed % on each node. Note that after a large number of
% steps, the PageRanks (in percentages) converge to the values given in Section
% 2.11 of Moler (alpha: .321, sigma: .2007, beta: .1705, delta: .1368,
% gamma: .1066, rho: .0643). See http://www.mathworks.com/moler for more
% details (the pagerank M-file, in particular).
%
% Note that this method is NOT how the PageRank is actually computed. Instead
% the eigenvalue problem A*x=x is solved for x, where A is the Markov
% transition matrix, A = p*G*D + e*z', where G is the binary matrix used here.
% The method here is a simplistic random-hopping demonstration of the Markov
% process, to motivate the A*x=x formulation of the problem. In this example,
% A does control how the transitions are made, but the matrix A is not formed
% explicitly.
%
% This demo only operates on a single graph. It is meant as a simple demo
% only, suitable for in-class use. To compute the PageRanks for an arbitrary
% graph, use pagerank.m, or the power method (repeat x=A*x until convergence,
% where A is the Markov transition matrix of the web).
%
% Example:
% pagerankdemo
% pagerankdemo (1000) % run 1000 steps with no user input, then quit
%
% See also pagerank
%
% I suggest single-stepping a dozen times or so to see the link traversal in
% process, and then type "1000". Hit control-C to quit.
%
% Copyright 2007, Tim Davis
% Initial graph
Graph = graphinit ;
rand ('state', 0) ;
n = size (Graph.G, 1) ;
help pagerankdemo
% initialize the page counts
hits = zeros (1,n) ;
oldwhere = 1 ;
where = 1 ;
hits (where) = 1 ;
set (Graph.node (where), 'FaceColor', [0 0 1]) ;
p = 0.85 ; % probability a link will be followed
c = sum (Graph.G) ; % outgoing degree
links = cell (1,n) ;
for k = 1:n
links {k} = find (Graph.G (:,k)) ;
end
follow_link = 0 ;
if (nargin < 1)
input ('hit enter to start at node alpha: ') ;
end
% write the stats to the figure
set (Graph.nodelabel (where), 'string', ...
sprintf ('%s %d (%3.1f%%)', Graph.nodes {where}, hits (where), ...
100 * hits (where) / sum (hits))) ;
if (nargin < 1)
input ('hit enter to take one step: ') ;
steps = 1 ;
end
% repeat
while (1)
% clear the old color and old arrow
set (Graph.node (where), 'FaceColor', [0 1 0]) ;
if (follow_link)
set (Graph.arrows (where,oldwhere), 'LineWidth', 2) ;
set (Graph.arrows (where,oldwhere), 'Color', [0 0 0]) ;
end
% determine where to go to next
oldwhere = where ;
if (c (where) == 0 || rand > p)
% no outgoing links, or ignore the links
follow_link = 0 ;
where = floor (n * rand + 1) ;
set (Graph.node (where), 'FaceColor', [0 0 1]) ;
else
% move along the link
follow_link = 1 ;
where = links{where}(floor (c (where) * rand + 1)) ;
set (Graph.node (where), 'FaceColor', [1 0 0]) ;
set (Graph.arrows (where,oldwhere), 'LineWidth', 5) ;
set (Graph.arrows (where,oldwhere), 'Color', [1 0 0]) ;
end
% increment the hit count
hits (where) = hits (where) + 1 ;
% write the stats to the figure
for k = 1:n
set (Graph.nodelabel (k), 'string', ...
sprintf ('%s %d (%3.1f%%)', Graph.nodes {k}, hits (k), ...
100 * hits (k) / sum (hits))) ;
end
drawnow
% go the next step
steps = steps - 1 ;
if (steps <= 0)
if (nargin > 0)
break ;
end
steps = input ...
('number of steps to make (default 1, control-C to quit): ') ;
if (steps == 0)
break ;
end
if (isempty (steps))
steps = 1 ;
end
end
end
%-------------------------------------------------------------------------------
function Graph = graphinit
% GRAPHINIT create the tiny-web example in Moler, section 2.11, and draw it.
% Example
% G = graphinit ;
figure (1)
clf
nodes = { 'alpha', 'beta', 'gamma', 'delta', 'rho', 'sigma' } ;
xy = [
0 4
1 3
1 2
2 4
2 0
0 0
] ;
x = xy (:,1) ;
y = xy (:,2) ;
% scale x and y to be in the range 0.1 to 0.9
x = 0.8 * x / 2 + .1 ;
y = 0.8 * y / 4 + .1 ;
xy = [x y] ;
xy_delta = [
.08 .04 0
-.03 -.02 -1
.04 0 0
-.05 .04 -1
-.03 0 -1
.03 0 0
] ;
xd = xy_delta (:,1) ;
yd = xy_delta (:,2) ;
tjust = xy_delta (:,3) ;
G = [
0 0 0 1 0 1
1 0 0 0 0 0
0 1 0 0 0 0
0 1 1 0 0 0
0 0 1 0 0 0
1 0 1 0 0 0 ] ;
clf
n = size (G,1) ;
axes ('Position', [0 0 1 1], 'Visible', 'off') ;
node = zeros (n,1) ;
nodelabel = zeros (n,1) ;
for k = 1:n
node (k) = annotation ('ellipse', [x(k)-.025 y(k)-.025 .05 .05]) ;
set (node (k), 'LineWidth', 2) ;
set (node (k), 'FaceColor', [0 1 0]) ;
nodelabel (k) = text (x (k) + xd (k), y (k) + yd (k), nodes {k}, ...
'Units', 'normalized', 'FontSize', 16) ;
if (tjust (k) < 0)
set (nodelabel (k), 'HorizontalAlignment', 'right') ;
end
end
axis off
% Yes, I realize that this is overkill; arrows should be sparse.
% This example is not meant for large graphs.
arrows = zeros (n,n) ;
[i j] = find (G) ;
for k = 1:length (i)
% get the center of the two nodes
figx = [x(j(k)) x(i(k))] ;
figy = [y(j(k)) y(i(k))] ;
% [figx figy] = dsxy2figxy (gca, axx, axy);
% shorten the arrows by s units at each end
s = 0.03 ;
len = sqrt (diff (figx)^2 + diff (figy)^2) ;
fy (1) = diff (figy) * (s/len) + figy(1) ;
fy (2) = diff (figy) * (1-s/len) + figy(1) ;
fx (1) = diff (figx) * (s/len) + figx(1) ;
fx (2) = diff (figx) * (1-s/len) + figx(1) ;
arrows (i(k),j(k)) = annotation ('arrow', fx, fy) ;
set (arrows (i(k),j(k)), 'LineWidth', 2) ;
set (arrows (i(k),j(k)), 'HeadLength', 20) ;
set (arrows (i(k),j(k)), 'HeadWidth', 20) ;
end
Graph.G = G ;
Graph.nodes = nodes ;
Graph.node = node ;
Graph.xy = xy ;
Graph.xy_delta = xy_delta ;
Graph.nodelabel = nodelabel ;
Graph.arrows = arrows ;
|