Hi,
I am a newbie at R and was thinking of ways to vectorize this loop but can't
quite figure out if anything more can be done :
Here the dimensions of the arrays are (N=2000, K=2, D=1000):
responsibilities : N * K
pi: K
corp: N*D
theta: K*D
for (n in 1:N)
{
for (k in 1:K)
{
responsibilities[n, k] <- (pi[k]) +
(log1p(dmultinom(x=corp[n, ], prob=exp(theta[k, ]))))
}
denom <- log.sum(responsibilities[n, ])
responsibilities[n,] <- responsibilities[n, ] - denom
}
Any help is much appreciated!
thanks.
[[alternative HTML version deleted]]