Forgetting by Pruning: Data Deletion in Join Cardinality Estimation

TL;DR

This paper introduces CEP, a novel unlearning framework for multi-table learned cardinality estimation systems, effectively handling data deletion with minimal retraining and high accuracy.

Contribution

CEP is the first framework tailored for data deletion in multi-table learned CE models, utilizing distribution sensitivity and domain pruning techniques.

Findings

CEP achieves the lowest Q-error in multi-table scenarios.

CEP outperforms full retraining under high deletion ratios.

CEP reduces convergence iterations with minimal computational overhead.

Abstract

Machine unlearning in learned cardinality estimation (CE) systems presents unique challenges due to the complex distributional dependencies in multi-table relational data. Specifically, data deletion, a core component of machine unlearning, faces three critical challenges in learned CE models: attribute-level sensitivity, inter-table propagation and domain disappearance leading to severe overestimation in multi-way joins. We propose Cardinality Estimation Pruning (CEP), the first unlearning framework specifically designed for multi-table learned CE systems. CEP introduces Distribution Sensitivity Pruning, which constructs semi-join deletion results and computes sensitivity scores to guide parameter pruning, and Domain Pruning, which removes support for value domains entirely eliminated by deletion. We evaluate CEP on state-of-the-art architectures NeuroCard and FACE across IMDB and TPC-H datasets. Results demonstrate CEP consistently achieves the lowest Q-error in multi-table scenarios, particularly under high deletion ratios, often outperforming full retraining. Furthermore, CEP significantly reduces convergence iterations, incurring negligible computational overhead of 0.3%-2.5% of fine-tuning time.