Suppose we have similarity values between some data point in the interval $[0, 1]$. How can I transform this similarity values into a dissimilarity values in the interval $[0, ∞]$?

  • $\begingroup$ Would simply taking the inverse work? This would give you values in the range 1 to infinity, so if you want the lower bound to be zero you can simply subtract 1. $\endgroup$
    – Oxbowerce
    Dec 28, 2021 at 17:39
  • $\begingroup$ I find it strange that you want to convert a normalized value to a non-normalized one. Why not just use the normalized dissimilarity $1-x$? $\endgroup$
    – Erwan
    Dec 29, 2021 at 0:07

1 Answer 1


You can use $-\ln x$ as transformation. It will map the interval $(0, 1]$ to the interval $[0, \infty)$.

As proposed in the comments it would be much easier to convert the normalized similarity value $x$ to a dissimilarity value by $1-x$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.