DiffPrep: Differentiable Data Preprocessing Pipeline Search for Learning over Tabular Data

Summary: DiffPrep introduces differentiable bi-level search for preprocessing pipelines, enabling optimization in a continuous space. Continuous relaxation enables descent to find pipelines with a single training, yielding up to 6.6pp gains on 15/18 datasets. (summarized by gpt-5-nano on Feb 09 2026)

Paper ID: h96fa0bac4f2d9267
Venue: SIGMOD
Year: 2023
Pagerank: 6.5621504e-05
Overall Rank: 4,524 | 69.59%
DOI: 10.1145/3589328

Incoming Non-self Citations Over Time

Authors

1. Peng Li (Georgia Institute of Technology)
2. Zhiyi Chen (Georgia Institute of Technology)
3. Xu Chu (Georgia Institute of Technology)
4. Kexin Rong (Georgia Institute of Technology)

BibTeX Citation

@inproceedings{li_sigmod23,
        title = {{DiffPrep: Differentiable Data Preprocessing Pipeline Search for Learning over Tabular Data}},
        author = {Li, Peng and Chen, Zhiyi and Chu, Xu and Rong, Kexin},
        series = {{SIGMOD} '23},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3589328},
        url = {https://dl.acm.org/doi/10.1145/3589328},
        year = {2023}
}

Incoming Citations (Sorted by Pagerank)

Showing 5 of 5 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank
6,921	CtxPipe: Context-aware Data Preparation Pipeline Construction for Machine Learning	2024	SIGMOD	5.6432616e-05
10,515	Understanding the Impact of Data Noise in Federated Learning: [Experiments & Analysis]	2026	SIGMOD	4.9793485e-05
10,990	MAPPipe: A System for Bridging Efficiency and Quality in Data Preprocessing via Knowledge-Augmented Structural Pruning	2026	VLDB	4.9793485e-05
11,061	LLM-AutoDP: Automatic Data Processing via LLM Agents for Model Fine-tuning	2026	VLDB	4.9793485e-05
11,368	Stress-Testing ML Pipelines with Adversarial Data Corruption	2025	VLDB	4.9793485e-05

Outgoing Citations (Sorted by Pagerank)

Showing 7 of 7 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
483	ActiveClean: Interactive Data Cleaning For Statistical Modeling	2016	VLDB	0.00017590977
975	Democratizing Data Science through Interactive Curation of ML Pipelines	2019	SIGMOD	0.00012750518
1,043	Data Cleaning: Overview and Emerging Challenges	2016	SIGMOD	0.00012335114
1,344	Detecting Data Errors: Where are we and what needs to be done?	2016	VLDB	0.00010956518
1,391	Creating Embeddings of Heterogeneous Relational Datasets for Data Integration Tasks	2020	SIGMOD	0.00010816237
1,847	Nearest Neighbor Classifiers over Incomplete Information: From Certain Answers to Certain Predictions	2021	VLDB	9.5120573e-05
3,617	Leva: Boosting Machine Learning Performance with Relational Embedding Data Augmentation	2022	SIGMOD	7.1574349e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	7,526	Auto-Prep: Holistic Prediction of Data Preparation Steps for Self-Service Business Intelligence	2025	VLDB
2	6,156	Automatic Data Acquisition for Deep Learning	2021	VLDB
3	12,049	Active Reinforcement Learning for Data Preparation: Learn2Clean with Human-In-The-Loop	2020	CIDR
4	6,921	CtxPipe: Context-aware Data Preparation Pipeline Construction for Machine Learning	2024	SIGMOD
5	6,028	A Scalable AutoML Approach Based on Graph Neural Networks	2022	VLDB
6	5,848	HAIPipe: Combining Human-generated and Machine-generated Pipelines for Data Preparation	2023	SIGMOD
7	4,235	Auto-Pipeline: Synthesizing Complex Data Pipelines By-Target Using Reinforcement Learning and Search	2021	VLDB
8	3,601	Where Is My Training Bottleneck? Hidden Trade-Offs in Deep Learning Preprocessing Pipelines	2022	SIGMOD
9	10,869	DeepPrep: An LLM-Powered Agentic System for Autonomous Data Preparation	2026	VLDB
10	10,990	MAPPipe: A System for Bridging Efficiency and Quality in Data Preprocessing via Knowledge-Augmented Structural Pruning	2026	VLDB