If we used log(#occurence) then the sparse count matrix will have lots of undefined log(0)'s no?
Also, I don't understand why the omega is the non-zero entries. This means we consider all entries except when count is 1?
I think Omega considering the non-zero entries means that only the elements of X (co-occurance matrix) that are not zero are considered. Example: elements that co-occure 1 time or more.
This assures that the log of the co-occurance is greater or equal to 0 (at least log(1)=0)
thanks it would make more sense indeed,
however in the lecture notes it defines 'count marix X' as the log of co-occurence matrix (as seen in the screenshot above).Where xdn=log(cooccurence(wd,wn)). So the way it is written in the lecture notes doesn't make sense since it asks for non zero entries of log(n) which doesn't seem right?
log of count for co-occurence matrix
Hello,
If we used log(#occurence) then the sparse count matrix will have lots of undefined log(0)'s no?
Also, I don't understand why the omega is the non-zero entries. This means we consider all entries except when count is 1?
Thanks for the help
2
Great question!
I think Omega considering the non-zero entries means that only the elements of X (co-occurance matrix) that are not zero are considered. Example: elements that co-occure 1 time or more.
This assures that the log of the co-occurance is greater or equal to 0 (at least log(1)=0)
I hope this helps :))
2
thanks it would make more sense indeed,
however in the lecture notes it defines 'count marix X' as the log of co-occurence matrix (as seen in the screenshot above).Where xdn=log(cooccurence(wd,wn)). So the way it is written in the lecture notes doesn't make sense since it asks for non zero entries of log(n) which doesn't seem right?
2
Add comment