refine mix rnd: done

chapter09/demo.m

@@ -40,18 +40,18 @@
 % subplot(2,1,1);plot(x); axis([x_range,y_range]); title('Original Signal');
 % subplot(2,1,2);plot(m); axis([x_range,y_range]); title('Recovery Signal');
 %% classification
-clear; close all
-k = 2;
-d = 2;
-n = 1000;
-[X,t] = kmeansRnd(d,k,n);
-[x1,x2] = meshgrid(linspace(min(X(1,:)),max(X(1,:)),n), linspace(min(X(2,:)),max(X(2,:)),n));
-
-[model, llh] = rvmBinEm(X,t-1);
-plot(llh);
-y = rvmBinPred(model,X)+1;
-figure;
-binPlot(model,X,y);
+% clear; close all
+% k = 2;
+% d = 2;
+% n = 1000;
+% [X,t] = kmeansRnd(d,k,n);
+% [x1,x2] = meshgrid(linspace(min(X(1,:)),max(X(1,:)),n), linspace(min(X(2,:)),max(X(2,:)),n));
+% 
+% [model, llh] = rvmBinEm(X,t-1);
+% plot(llh);
+% y = rvmBinPred(model,X)+1;
+% figure;
+% binPlot(model,X,y);
 %% demo: kmeans 
 % close all; clear;
 % d = 2;
@@ -98,4 +98,4 @@
 % figure;
 % plotClass(X,z);
 % figure;
-% plot(llh);
+% plot(llh);

chapter09/kmeansRnd.m

@@ -2,11 +2,11 @@
 % Sampling from a Gaussian mixture distribution with common variances (kmeans model).
 % Written by Michael Chen ([email protected]).
 alpha = 1;
-beta = 6*nthroot(k,d);
+beta = k;
 
 X = randn(d,n);
 w = dirichletRnd(alpha,ones(1,k)/k);
 z = discreteRnd(w,n);
 E = full(sparse(z,1:n,1,k,n,n));
-center = rand(d,k)*beta;
+center = randn(d,k)*beta;
 X = X+center*E;

chapter09/mixGaussRnd.m

@@ -5,7 +5,7 @@
 W0 = eye(d);  % hyperparameter of inverse Wishart prior of covariances
 v0 = d+1;  % hyperparameter of inverse Wishart prior of covariances
 mu0 = zeros(d,1);  % hyperparameter of Guassian prior of means
-beta0 = 1/(nthroot(k,d))^2; % hyperparameter of Guassian prior of means
+beta0 = k; % hyperparameter of Guassian prior of means
 
 w = dirichletRnd(alpha0,ones(1,k)/k);
 z = discreteRnd(w,n);
@@ -16,7 +16,7 @@
 for i = 1:k
     idx = z==i;
     Sigma(:,:,i) = iwishrnd(W0,v0); % invpd(wishrnd(W0,v0));
-    mu(:,i) = gaussRnd(mu0,Sigma(:,:,i)/beta0);
+    mu(:,i) = gaussRnd(mu0,beta0*Sigma(:,:,i));
     X(:,idx) = gaussRnd(mu(:,i),Sigma(:,:,i),sum(idx));
 end
 model.mu = mu;