You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What is the bug?
The formula for the innerproduct distance measure equals the l1 distance metric, not dotproduct. If I get it correctly, innerproduct in scriptscoring is fully supported and correctly uses the dotproduct formula as distance metric (see this PR in opendistro). So this formula should be corrected in order to avoid confusion. Otherwise (if it's actually calculating l1 for innerproduct space_type) there should be an explanation why this is the case.
How can one reproduce the bug?
Steps to reproduce the behavior:
tstadel
changed the title
[BUG] correct distance metric formula for Innerproduct in scriptscoring
[BUG] correct distance metric formula for innerproduct in scriptscoring
Feb 7, 2022
tstadel
changed the title
[BUG] correct distance metric formula for innerproduct in scriptscoring
[BUG] incorrect distance metric formula for innerproduct in scriptscoring
Feb 7, 2022
tstadel
changed the title
[BUG] incorrect distance metric formula for innerproduct in scriptscoring
[BUG] incorrect distance metric formula for innerproduct in knn scriptscoring
Feb 7, 2022
What is the bug?
The formula for the innerproduct distance measure equals the l1 distance metric, not dotproduct. If I get it correctly, innerproduct in scriptscoring is fully supported and correctly uses the dotproduct formula as distance metric (see this PR in opendistro). So this formula should be corrected in order to avoid confusion. Otherwise (if it's actually calculating l1 for innerproduct space_type) there should be an explanation why this is the case.
How can one reproduce the bug?
Steps to reproduce the behavior:
What is the expected behavior?
Innerproduct formula should equal the dotproduct formula.
The text was updated successfully, but these errors were encountered: