Re: Computational complexity of finding MLEs?

But how do we know that we even need to compute derivatives of the
log-likelihood function? For instance, we can estimate the mean
parameter of the Normal distribution by just averaging the observations
together. As opposed to logistic regression, where we may have to
compute derivatives and iterate several times. Are we able to rule out
the existence of a a special purpose efficient algorithm (like for
estimating mean of the Normal distribution) for cases where Newton-type
methods are typically used?

.

Loading