What does it mean to be SET distance/similarity metric?

47 Views Asked by At

I was looking at Tverski contrast distance :

d(x,y) = a * f(x intersect y) - b * f(x-y) - c * f(y-x)

in binary :

d(x,y) = a * count(x&y) - b * count(x&~y) - c * count(~x&y)

It is said that it is SET DISTANCE metric !!

How to understand this ? Beside the obvious, what does it mean to be a SET metric. For distance like Euclidean and Hamming it is easier to visualize it as distance between two points in N-dimentional space, not so for a SET !