Towards Practical Learned Indexing

TL;DR

This paper introduces PLEX, a practical learned index structure that simplifies hyperparameter tuning, guarantees error bounds, and improves build and lookup efficiency by combining splines and radix layers.

Contribution

PLEX is a new learned index that uses only one hyperparameter, offers error guarantees, and balances build and query performance better than existing methods.

Findings

PLEX reduces hyperparameter complexity to a single parameter.

PLEX achieves better build and lookup trade-offs than prior learned indexes.

PLEX guarantees maximum prediction error, enhancing reliability.

Abstract

Latest research proposes to replace existing index structures with learned models. However, current learned indexes tend to have many hyperparameters, often do not provide any error guarantees, and are expensive to build. We introduce Practical Learned Index (PLEX). PLEX only has a single hyperparameter $\epsilon$ (maximum prediction error) and offers a better trade-off between build and lookup time than state-of-the-art approaches. Similar to RadixSpline, PLEX consists of a spline and a (multi-level) radix layer. It first builds a spline satisfying the given $\epsilon$ and then performs an ad-hoc analysis of the distribution of spline points to quickly tune the radix layer.