Lune

ICLR2025Top-tier venue

Beyond the convexity assumption: Realistic tabular data generation under quantifier-free real linear constraints

Mihaela C. Stoian, Eleonora Giunchiglia

Abstract

Synthetic tabular data generation has traditionally been a challenging problem due to the high complexity of the underlying distributions that characterise this type of data. Despite recent advances in deep generative models (DGMs), existing methods often fail to produce realistic datapoints that are well-aligned with available background knowledge. In this paper, we address this limitation by introducing Disjunctive Refinement Layer (DRL), a novel layer designed to enforce the alignment of generated data with the background knowledge specified in user-defined constraints. DRL is the first method able to automatically make deep learning models inherently compliant with constraints as expressive as quantifier-free linear formulas, which can define non-convex and even disconnected spaces. Our experimental analysis shows that DRL not only guarantees constraint satisfaction but also improves efficacy in downstream tasks. Notably, when applied to DGMs that frequently violate constraints, DRL eliminates violations entirely. Further, it improves performance metrics by up to 21.4% in F1-score and 20.9% in Area Under the ROC Curve, thus demonstrating its practical impact on data generation.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

7 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • A-NeSI: A Scalable Approximate Method for Probabilistic Neurosymbolic Inference cites Semantic Probabilistic Layers for Neuro-Symbolic Learning
  • Semantic Probabilistic Layers for Neuro-Symbolic Learning cites MultiplexNet: Towards Fully Satisfied Logical Constraints in Neural Networks
  • The Theory and Practice of MAP Inference over Non-Convex Constraints cites Neurosymbolic Diffusion Models
  • Doubling Your Data in Minutes: Ultra-fast Tabular Data Generation via LLM-Induced Dependency Graphs cites TabDDPM: Modelling Tabular Data with Diffusion Models
  • Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI cites A-NeSI: A Scalable Approximate Method for Probabilistic Neurosymbolic Inference
  • Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI cites Semantic Probabilistic Layers for Neuro-Symbolic Learning
  • Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI cites MultiplexNet: Towards Fully Satisfied Logical Constraints in Neural Networks
  • Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI cites Neurosymbolic Diffusion Models
  • Neurosymbolic Diffusion Models cites A-NeSI: A Scalable Approximate Method for Probabilistic Neurosymbolic Inference
  • Neurosymbolic Diffusion Models cites Semantic Probabilistic Layers for Neuro-Symbolic Learning
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers7

  1. Neurosymbolic Diffusion ModelsEmile van Krieken, Pasquale Minervini, Edoardo Maria Ponti, Antonio VergariNeurIPS 2025 · 12 citations
  2. Doubling Your Data in Minutes: Ultra-fast Tabular Data Generation via LLM-Induced Dependency GraphsShuo Yang, Zheyu Zhang, Bardh Prenkaj, Gjergji KasneciEMNLP 2025 · 6 citations
  3. Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AILuca Andolfi, Eleonora GiunchigliaNeurIPS 2025 · 4 citations
  4. Synthesising Counterfactual Explanations via Label-Conditional Gaussian Mixture Variational AutoencodersJunqi Jiang, Francesco Leofante, Antonio Rago, Francesca ToniICLR 2026 · 2 citations
  5. The Theory and Practice of MAP Inference over Non-Convex ConstraintsLeander Kurscheidt, Gabriele Masina, Roberto Sebastiani, Antonio VergariICML 2026 · 1 citation

Builds on14

  1. TabDDPM: Modelling Tabular Data with Diffusion ModelsAkim Kotelnikov, Dmitry Baranchuk, Ivan Rubachev, Artem BabenkoICML 2023 · 518 citations
  2. DECAF: Generating Fair Synthetic Data Using Causally-Aware Generative NetworksBoris van Breugel, Trent Kyono, Jeroen Berrevoets, Mihaela van der SchaarNeurIPS 2021 · 174 citations
  3. Semantic Probabilistic Layers for Neuro-Symbolic LearningKareem Ahmed, Stefano Teso, Kai-Wei Chang, Guy Van den Broeck et al.NeurIPS 2022 · 133 citations
  4. MultiplexNet: Towards Fully Satisfied Logical Constraints in Neural NetworksNick Hoernle, Rafael-Michael Karampatsis, Vaishak Belle, Kobi GalAAAI 2022 · 73 citations
  5. A-NeSI: A Scalable Approximate Method for Probabilistic Neurosymbolic InferenceEmile van Krieken, Thiviyan Thanapalasingam, Jakub M. Tomczak, Frank van Harmelen et al.NeurIPS 2023 · 62 citations

Related papers

  1. How Realistic Is Your Synthetic Data? Constraining Deep Generative Models for Tabular DataMihaela C. Stoian, Salijona Dyrmishi, Maxime Cordy, Thomas Lukasiewicz et al.ICLR 2024 · 33 citations
  2. GOGGLE: Generative Modelling for Tabular Data by Learning Relational StructureTennison Liu, Zhaozhi Qian, Jeroen Berrevoets, Mihaela van der SchaarICLR 2023
  3. Learning with Logical Constraints but without Shortcut SatisfactionZenan Li, Zehua Liu, Yuan Yao, Jingwei Xu et al.ICLR 2023
  4. DisjunctiveNet: Neural Symbolic Learning via Differentiable Convexified Optimization LayersShraman Pal, Can LiICML 2026
  5. ConSequence: Synthesizing Logically Constrained Sequences for Electronic Health Record GenerationBrandon Theodorou, Shrusti Jain, Cao Xiao, Jimeng SunAAAI 2024
Dusk over the sea between two cliffs drawn in fine vertical lines