A Review for Weighted MinHash Algorithms

TL;DR

This paper reviews various weighted MinHash algorithms, categorizes their types, discusses their evolution, and provides a Python toolbox for experimental comparison with standard MinHash methods.

Contribution

It offers a comprehensive categorization and discussion of weighted MinHash algorithms, including their evolution and connections, along with a practical Python toolbox for experimentation.

Findings

Weighted MinHash algorithms are categorized into quantization-based, active index-based, and others.

The evolution from integer to real-valued weighted MinHash algorithms is analyzed.

Experimental comparison highlights differences between standard MinHash and weighted MinHash methods.

Abstract

Data similarity (or distance) computation is a fundamental research topic which underpins many high-level applications based on similarity measures in machine learning and data mining. However, in large-scale real-world scenarios, the exact similarity computation has become daunting due to "3V" nature (volume, velocity and variety) of big data. In such cases, the hashing techniques have been verified to efficiently conduct similarity estimation in terms of both theory and practice. Currently, MinHash is a popular technique for efficiently estimating the Jaccard similarity of binary sets and furthermore, weighted MinHash is generalized to estimate the generalized Jaccard similarity of weighted sets. This review focuses on categorizing and discussing the existing works of weighted MinHash algorithms. In this review, we mainly categorize the Weighted MinHash algorithms into quantization-based approaches, "active index"-based ones and others, and show the evolution and inherent connection of the weighted MinHash algorithms, from the integer weighted MinHash algorithms to real-valued weighted MinHash ones (particularly the Consistent Weighted Sampling scheme). Also, we have developed a python toolbox for the algorithms, and released it in our github. Based on the toolbox, we experimentally conduct a comprehensive comparative study of the standard MinHash algorithm and the weighted MinHash ones.