Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Return the same input vector if its a unit vector in VectorUtil#l2normalize (#12726) 

### Description

While going through [VectorUtil](https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/util/VectorUtil.java) class, I observed we don't have a check for unit vector in `VectorUtil#l2normalize` so passing a unit vector goes thorough the whole L2 normalization(which is totally not required and it should early exit?). I confirmed this by trying out a silly example of `VectorUtil.l2normalize(VectorUtil.l2normalize(nonUnitVector))` and it performed every calculation twice. We could also argue that user should not call for this for a unit vector but I believe there would be cases where user simply want to perform the L2 normalization without checking the vector or if there are some overflowing values.

TL;DR : We should early exit in `VectorUtil#l2normalize`, returning the same input vector if its a unit vector

This is easily avoidable if we introduce a light check to see if the L1 norm or squared sum of input vector is equal to 1.0 (or) maybe just check `Math.abs(l1norm - 1.0d) <= 1e-5` (as in this PR) because that unit vector dot product(`v x v`) are not exactly 1.0 but like example : `0.9999999403953552` etc. With `1e-5` delta here we would be assuming a vector v having `v x v` >= `0.99999` is a unit vector or say already L2 normalized which seems fine as the delta is really small? and also the check is not heavy one?.
This commit is contained in:
Shubham Chaudhary 2023-10-30 19:22:28 +05:30 committed by GitHub
parent 2a8d187a99
commit 7943b7ad1c
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 10 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lucene/CHANGES.txt
View File

@ -242,6 +242,8 @@ Optimizations
* GITHUB#12702: Disable suffix sharing for block tree index, making writing the terms dictionary index faster * GITHUB#12702: Disable suffix sharing for block tree index, making writing the terms dictionary index faster
and less RAM hungry, while making the index a bit (~1.X% for the terms index file on wikipedia). (Guo Feng, Mike McCandless) and less RAM hungry, while making the index a bit (~1.X% for the terms index file on wikipedia). (Guo Feng, Mike McCandless)
* GITHUB#12726: Return the same input vector if its a unit vector in VectorUtil#l2normalize. (Shubham Chaudhary)
Changes in runtime behavior Changes in runtime behavior
--------------------- ---------------------

13
lucene/core/src/java/org/apache/lucene/util/VectorUtil.java
View File

@ -106,18 +106,21 @@ public final class VectorUtil {
* @throws IllegalArgumentException when the vector is all zero and throwOnZero is true * @throws IllegalArgumentException when the vector is all zero and throwOnZero is true
*/ */
public static float[] l2normalize(float[] v, boolean throwOnZero) { public static float[] l2normalize(float[] v, boolean throwOnZero) {
double squareSum = IMPL.dotProduct(v, v); double l1norm = IMPL.dotProduct(v, v);
int dim = v.length; if (l1norm == 0) {
if (squareSum == 0) {
if (throwOnZero) { if (throwOnZero) {
throw new IllegalArgumentException("Cannot normalize a zero-length vector"); throw new IllegalArgumentException("Cannot normalize a zero-length vector");
} else { } else {
return v; return v;
} }
} }
double length = Math.sqrt(squareSum); if (Math.abs(l1norm - 1.0d) <= 1e-5) {
return v;
}
int dim = v.length;
double l2norm = Math.sqrt(l1norm);
for (int i = 0; i < dim; i++) { for (int i = 0; i < dim; i++) {
v[i] /= length; v[i] /= (float) l2norm;
} }
return v; return v;
} }