krzysztof/landing_page
1
0
Fork 0
mirror of https://github.com/qdrant/landing_page.git synced 2026-10-09 21:08:31 +02:00
Code Issues Packages Projects Releases Wiki Activity

review remark

Browse Source
This commit is contained in:
Ivan Pleshkov
2026-05-11 22:29:43 +02:00
parent 1cf25ad0d9
commit d91e94a365
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
qdrant-landing/content/articles/turboquant-quantization.md
+1 -1
View File
@@ -136,7 +136,7 @@ Truly isotropic data matches the theoretical Gaussian quantiles, the formula col
### L2 and Unnormalized Dot
Vanilla TurboQuant assumes all inputs live on the unit sphere — that is, cosine distance only. We extend the algorithm beyond the sphere by **storing the original L2 norm** and restoring L2 and unnormalized dot from normalized one, so dot and L2 cost the same as cosine in the hot path.
Vanilla TurboQuant assumes all inputs live on the unit sphere — that is, cosine distance only. We extend the scoring mechanism and unlock L2 and unnormalized dot by *storing the original L2 norm*, normalizing the vectors and then apply the L2 norm back during scoring.
L2 distances are reconstructed via the identity `‖q − v‖² = ‖q‖² + ‖v‖² − 2⟨q, v⟩ = ‖q‖² + ‖v‖² − 2 ‖v‖ ‖q‖ ⟨q_normalized, v_normalized⟩`, where all components on the right-hand side are already available.