The negative weights made the distance equation quite complicated and difficult to parse. I suggest removing them, because they aren't used in the front end. I'm also not confident they're being calculated correctly. In the equation below, the +2 here makes the range different for the negative weights compared to positive weights (since all values are in [-1, 1]). If we don't remove them, please:
A) in the documentation, first describe the main simple positive weight case, then add the negative weight case as a separate section (so as not to confuse users who won't use the negative weights)
The negative weights made the distance equation quite complicated and difficult to parse. I suggest removing them, because they aren't used in the front end. I'm also not confident they're being calculated correctly. In the equation below, the
+2
here makes the range different for the negative weights compared to positive weights (since all values are in [-1, 1]). If we don't remove them, please:A) in the documentation, first describe the main simple positive weight case, then add the negative weight case as a separate section (so as not to confuse users who won't use the negative weights)
B) double check that the equations are correct
screenshot from similarity_conceptual.ipynb: