Research map / Image generation / Fast diffusion sampling
Towards Synthesizing High-Dimensional Tabular Data with Limited Samples
Zuqing Li, Junhao Gan, Jianzhong Qi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add noise-conditioned control module to TabSyn | TabSyn · GE | F1: 0.450 -> 0.635 | Improved |
| add control module to TabSyn | TabSyn · UG regression | RMSE: 0.267 -> 0.069 | Improved |
| add control module to TabSyn | TabSyn · ED | AUC: 0.952 -> 0.988 (SMOTE 0.990) | Mixed |
| use staged training with control module vs joint training (ablation) | CtrTab · 7 datasets avg | F1/R2: 0.772 -> 0.868 | Improved |
| add last fusion connection to denoising network output (ablation) | CtrTab · 7 datasets avg | F1/R2: 0.835 -> 0.868 | Improved |
| use control module instead of simply feeding noisy x0 to TabSyn (ablation) | TabSyn · 7 datasets avg | F1/R2: 0.751 -> 0.868 | Improved |
| use control module instead of doubling training epochs (ablation) | TabSyn · 7 datasets avg | F1/R2: 0.746 -> 0.868 | Improved |
| use control module instead of doubling model size with noisy data duplication (ablation) | TabSyn · 7 datasets avg | F1/R2: 0.822 -> 0.868 | Improved |
| use control module instead of dropout regularization (ablation) | TabSyn · 7 datasets avg | F1/R2: 0.707 -> 0.868 | Improved |
| add small Laplace noise to control signal (vs zero noise) (ablation) | CtrTab · noise scale study | F1 and R2 | Improved |
| increase Laplace noise scale to very large values (ablation) | CtrTab · noise scale study | F1 and R2 | Hurt |
| replace Laplace noise with Gaussian or uniform noise (ablation) | CtrTab · noise type | accuracy | No clear effect |
Closest papers by meaning
- TabDiff: a Mixed-type Diffusion Model for Tabular Data Generation (ICLR 2025 · similarity 0.88)
- CTSyn: A Foundation Model for Cross Tabular Data Generation (ICLR 2025 · similarity 0.88)
- TabNAT: A Continuous-Discrete Joint Generative Framework for Tabular Data (ICML 2025 · similarity 0.88)
- Mixed-Type Tabular Data Synthesis with Score-based Diffusion in Latent Space (ICLR 2024 · similarity 0.88)
- Continuous Diffusion for Mixed-Type Tabular Data (ICLR 2025 · similarity 0.88)
- ClavaDDPM: Multi-relational Data Synthesis with Cluster-guided Diffusion Models (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
