SAmix {mcsm} | R Documentation |
This function implements a simulated annealing algorithm to optimize the posterior distribution of a normal mixture with two components and only the means unknown,
like=function(mu){
-sum(log((.25*dnorm(da-mu[1])+.75*dnorm(da-mu[2]))))}
with a schedule temp=1/log(1+t).
SAmix(x, tolerance = 10^(-4), factor = 1)
x |
two-dimensional vector, starting point of the simulated annealing algorithm |
tolerance |
maximal difference in the target value needed to stop the simulated annealing algorithm |
factor |
scale factor of temp that determines the scale of the random walk |
theta |
sequence of points explored by the simulated annealing algorithm |
like |
corresponding sequence of posterior values |
ite |
number of iterations to reach stable value |
Christian P. Robert and George Casella
From Chapter 5 of EnteR Monte Carlo Statistical Methods
da=sample(rbind(rnorm(10^2),2.5+rnorm(3*10^2))) SAres=SAmix(x=c(-.3,.6),tol=10^(-2),fac=.1)