Dear all
to extract the most relevant input variables to an output one, I have 
used MUTUAL INFORMATION BASED SHANON ENTROPY.

my question

can the correlation measure do the same thing, or it just detect the 
linear dependancy without detecting the relevance of input variables to 
an output one?

I give an example, the selling of ice crean and solar glasses in
summer is highly correlated but we can't say that one variable is
relevant (explain) to other one. if we use mutual information can we
detect this irrelevance (or relevance) beween these two variables.  in
my case can I say that mutual information is better than correlation
to detect the relevance of input variables to one output

Many thanks in advance
faten


Reply via email to