Feature vector binarization

WO2024264068A3 Publication Date: 2025-03-27SILVERARROW AI INC
2 Cites 0 Cited by

Patent Information

Application Number
PCT/US2024/035313
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-06-23
Filing Date
2024-06-24
Publication Date
2025-03-27

AI Technical Summary

Technical Problem

Conventional feature vector binarization techniques fail to preserve the relative distance relationships between non-neighboring quantization values, leading to reduced comparison accuracy and performance, especially when using Hamming distance, which results in significant loss of entropy and accuracy in similarity measurements.

Method used

The method involves transforming and quantizing feature vectors to generate binarized feature vectors with independent bits, preserving actual distances and allowing for better Hamming distance comparisons by projecting features onto additional coordinate systems and using multiple transforms to recover lost entropy.

Benefits of technology

This approach improves computer processing speed, reduces power consumption, and enhances the overall performance of artificial intelligence systems by maintaining accurate similarity measurements and reducing the complexity of feature comparisons.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

A method includes identifying a feature vector with N features to transform into M bits. The feature vector including one or more values, where each value corresponds to a characteristic of an object. The method includes applying a transform to the feature vector to create a transformed feature vector. The method further includes quantizing the transformed feature vector to generate a binarized feature vector of M bits.
Need to check novelty before this filing date? Find Prior Art