Modeling Tabular data using Conditional GAN

By Lei Xu, Maria Skoularidou, Alfredo Cuesta-Infante, Kalyan Veeramachaneni. Neural Information Processing Systems, 2019

2,053 citations recorded by the source corpus.

Modeling the probability distribution of rows in tabular data and generating realistic synthetic data is a non-trivial task. Tabular data usually contains a mix of discrete and continuous columns. Continuous columns may have multiple modes whereas discrete columns are sometimes imbalanced making the modeling difficult. Existing statistical and deep neural network models fail to properly model this type of data. We design TGAN, which uses a conditional generative adversarial network to address these challenges. To aid in a fair and thorough comparison, we design a benchmark with 7 simulated and 8 real datasets and several Bayesian network baselines. TGAN outperforms Bayesian methods on most of the real datasets whereas other deep learning methods could not.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/e7f3478fd8aac6940a4bf4f5eb60ac38f6b0b85b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Generating synthetic data in finance: opportunities, challenges and pitfalls (2020, methodology): “At the time of writing, the most common method is to adapt classical statistical methods such as the Kolmogorov-Smirnov test [51, 52], although more recently some effort has gone towards developing specialised metrics [53].”
Language Models are Realistic Tabular Data Generators (2022, methodology): “Therefore, the generation of realistic artiﬁcial tabular data has received considerable attention in recent years (Choi et al., 2017; Park et al., 2018; Xu et al., 2019; Borisov et al., 2021).”
TabNet: Attentive Interpretable Tabular Learning (2019, background): “stacked convolutional layers or multi-layer perceptrons (MLPs) are vastly overparametrized – the lack of appropriate inductive bias often causes them to fail to find optimal solutions for tabular decision manifolds (Goodfellow, Bengio, and Courville 2016; Shavitt and Segal 2018; Xu et al. 2019).”
Fast, Accurate, and Simple Models for Tabular Data via Augmented Distillation (2020, background): “[44] for augmenting tabular data using conditional generative adversarial networks (GANs), which have been shown to produce higher-quality samples than other deep generative models.”
A review of ensemble learning and data augmentation models for class imbalanced problems: combination, implementation and evaluation (2023, background): “We note that one of the reasons that CT-GANs have not outperformed the established methods in the literature could be because CT-GANs [259] have been designed for tabular data that feature both discrete and continuous values.”
CTAB-GAN: Effective Table Data Synthesizing (2021, methodology): “Beyond its success on generating images, GAN has recently been applied to generate tabular data [16, 19, 24, 25].”
Conditional Wasserstein GAN-based Oversampling of Tabular Data for Imbalanced Learning (2020, background): “However, weight-clipping constrains the discriminator’s capacity, leads to the weights being pushed to the two extreme values of the permitted range [ − c, c ] and can lead to exploding or vanishing gradients.”
TabDDPM: Modelling Tabular Data with Diffusion Models (2022, methodology): “• CTGAN (Xu et al., 2019) — arguably the most popular and well-known GAN-based model for synthetic data generation.”
Deep Neural Networks and Tabular Data: A Survey (2021, methodology): “They also use the normalization proposed in [130].”
Breaking Distortion-free Watermarks in Large Language Models (2025, unknown): “…or introducing randomness during generation, (ii) extending testing attacks under stricter constraints (e.g., limited queries or partial detector access), (iii) scaling to larger models and (iv) exploring how spoofing interacts with other provenance mechanisms like model fingerprinting [28].”
TabDLM: Free-Form Tabular Data Generation via Joint Numerical-Language Diffusion (2026, unknown): “For baselines, we compare T AB DLM with some widely-used synthetic tabular data generation methods from four categories: 1) GAN-based: CTGAN (Xu et al., 2019); 2) VAE-based: TVAE (Xu et al., 2019) and GOGGLE (Liu et al., 2023); 3) Autore-gressive Tabular Language Model: GReaT (Borisov et al., 2023)…”