Fast Construction of Learned Count-Min Sketch via Ternary Search

2026-08-09

AI Summary

This work addresses the inefficiency of brute-force hyperparameter tuning in traditional Learned Count-Min Sketch, which lacks a systematic and efficient optimization strategy. The study introduces ternary search to this task for the first time, leveraging an analysis of the relationship between model performance and the number of hash functions. It proposes a novel approach that employs only a single hash function and rapidly identifies the optimal proportion of unique buckets. This method achieves estimation accuracy comparable to brute-force search while dramatically improving optimization efficiency—accelerating the process by 216–729× when model performance is strong and still yielding 221–740× speedups even under suboptimal conditions. The approach establishes a highly efficient new paradigm for weighted error optimization in frequency estimation.

Abstract

The Learned Count-Min Sketch (LCMS) is a learned data structure that estimates element frequencies in a multiset and has been experimentally shown to outperform classical data structures in the capacity-accuracy trade-off. However, its performance depends heavily on parameter selection. Because systematic optimization has not been adequately discussed, previous approaches relied on inefficient brute-force methods. In this study, we propose a method to rapidly optimize the parameters of the original LCMS. We experimentally confirmed that when the machine learning model performs well enough, using a single hash function is sufficient to optimize the weighted error metric. Based on this, we introduce a ternary search approach to efficiently find the optimal proportion of Unique Buckets. Our method achieves the same performance as brute-force approaches while speeding up parameter optimization by $216$-$729$ times when the machine learning model's performance is sufficient. Furthermore, even when the model's performance is suboptimal, our approach still determines appropriate parameters $221$-$740$ times faster than the brute-force approach.

Problem

• Learned Count-Min Sketch
• parameter optimization
• brute-force search
• frequency estimation
• data structures

Innovation

• Learned Count-Min Sketch
• ternary search
• parameter optimization
• frequency estimation
• learned data structures