diff toolboxes/distance_learning/mlr/mlr_plot.m @ 0:e9a9cd732c1e tip

first hg version after svn

--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/toolboxes/distance_learning/mlr/mlr_plot.m	Tue Feb 10 15:05:51 2015 +0000
@@ -0,0 +1,76 @@
+function mlr_plot(X, Y, W, D)
+
+    [d, n] = size(X);
+    %%%
+    % Color-coding
+    CODES = {   'b.', 'r.', 'g.', 'c.', 'k.', 'm.', 'y.', ...
+                'b+', 'r+', 'g+', 'c+', 'k+', 'm+', 'y+', ...
+                'b^', 'r^', 'g^', 'c^', 'k^', 'm^', 'y^', ...
+                'bx', 'rx', 'gx', 'cx', 'kx', 'mx', 'yx', ...
+                'bo', 'ro', 'go', 'co', 'ko', 'mo', 'yo'};
+
+    %%%
+    % First, PCA-plot of X
+
+    figure;
+
+    z = sum(X,3);
+    subplot(3,2,[1 3]), pcaplot(z, eye(d), Y, CODES), title('Native');
+    subplot(3,2,[2 4]), pcaplot(X, W, Y, CODES), title('Learned');
+
+    [vecs, vals] = eig(z * z');
+    subplot(3,2,5), bar(sort(real(diag(vals)), 'descend')), title('X*X'' spectrum'), axis tight;
+
+    if size(X,3) == 1
+        [vecs, vals] = eig(W);
+        vals = real(diag(vals));
+    else
+        if size(W,3) == 1
+            vals = real(W(:));
+        else
+            vals = [];
+            for i = 1:size(W,3)
+                [vecs, vals2] = eig(W(:,:,i));
+                vals = [vals ; real(diag(vals2))];
+            end
+        end
+    end
+    subplot(3,2,6), bar(sort(vals, 'descend')), title('W spectrum'), axis tight;
+
+    if nargin < 4
+        return;
+    end
+
+    %%%
+    % Now show some diagnostics
+
+    figure;
+    subplot(2,1,1), semilogy(D.f), title('Objective');
+    subplot(2,1,2), barh([D.time_SO, D.time_solver D.time_total]), ...
+                    title('Time% in SO/solver/total');
+
+
+function pcaplot(X, W, Y, CODES)
+    if size(X,3) == 1
+        A = X' * W * X;
+    else
+        A = 0;
+        if size(W,3) == 1
+            for i = 1:size(X,3)
+                A = A + X(:,:,i)' * bsxfun(@times, W(:,i), X(:,:,i));
+            end
+        else
+            for i = 1:size(X,3)
+                A = A + X(:,:,i)' * W(:,:,i) * X(:,:,i);
+            end
+        end
+    end
+    [v,d] = eigs(A, 3);
+    X2 = d.^0.5 * v';
+    hold on;
+    for y = 1:max(Y)
+        z = Y == y;
+        scatter3(X2(1, z), X2(2, z), X2(3,z), CODES{y});
+    end
+
+    axis equal;