let's say I calculate JSD between two distributions A and B as: JSD(A, B)= x and JSD(C, D)=y and JSD(K,L)= z. If x>y>z, can I say K and L are closer to each other than D and C and A and B? My question is not about what kind of closeness JSD is measuring, I simply want to know if JDS can be used to compare the distance of a pair of data distributions with another pair of data distributions in this example, you can imagine the (A, B) pair and (C,D) pair. Note that in this example, I am not comparing the JSD of A and C to the same reference data distribution.
Asked
Active
Viewed 10 times
