Re: weighted score

[Tamas Jambor]

On 24/02/2010 00:28, Ted Dunning wrote:

For many applications where averages seem usable with positive weights, I
would use squared distance from positive examples and negative squared
distance from negative examples.


   
so it is going to maximize the distance between the prediction and 
dissimilar users, and minimize otherwise.
is it something like this?

argmin sum(c*(p-r)^2)

where c is the correlation, p is the prediction and r is the true rating.


T