Math Forum Discussions

cagdas.ozgenc@gmail.com

Posts: 58
Registered: 3/29/06

Maximum Mutual Information vs Maximum Likelihood
Posted: Sep 29, 2010 8:01 AM

Plain Text

Greetings,

How would you compare Maximum Mutual Information Estimation to Maximum
Likelihood estimation? In which cases they yield equivalent results,
in which cases they diverge significantly?

If we look at the estimation of a linear regression

Y = aX + b

is it better to maximize
a) mutual information between Y and aX + b or
b) ML of Y - (aX+b)?

I understand it is not well defined what's meant by better here.
Nevertheless, I am just trying to understand under which conditions
one should be preferred over the other and some pitfalls to avoid.

Thanks in advance