Research map / Image generation / GANs
Double Generative Adversarial Networks for Conditional Independence Testing
Chengchun Shi, Tianlin Xu, Wicher Bergsma, Lexin Li
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 |
|---|---|---|---|
| use double GANs doubly-robust statistic instead of GCIT single generator | conditional independence test · Simulation, Z normal, d_Z=250, alpha=0.1 | empirical type-I error: 0.15 | Improved |
| use DGCIT instead of KCIT in high dimension | conditional independence test · Simulation, d_Z>=50 | empirical type-I error | Improved |
| use DGCIT instead of KCIT in low dimension | KCIT · Simulation, d_Z=5 | empirical type-I error | No clear effect |
| use DGCIT instead of GCIT and RCIT for power | conditional independence test · Simulation, b from 0.3 to 0.9, d_Z=100,200 | empirical power: power approaches 1 as b increases to 0.9 | Improved |
| use maximum over B transformation functions instead of a single generalized covariance (ablation) | DGCIT · design | test power | Improved |
| use MMD-type statistic on (X,Y,Z) instead of the (X,Y) maximum statistic (ablation) | DGCIT · Numerical experiments | test power | Hurt |
| increase number of transformation functions B from 30 to 50 (ablation) | DGCIT · implementation | power vs computation: B between 30 and 50 | No clear effect |
| vary pseudo samples M and splits L (ablation) | DGCIT · implementation | test result: M=100, L=3 | No clear effect |
| use DGCIT instead of GCIT on MAP3K5 and FTL3 mutations | conditional independence test · CCLE PLX4720 | p-value: MAP3K5 <0.001 -> 0.794; FTL3 0.521 -> 0 | Mixed |
Closest papers by meaning
- Conditional Independence Testing using Generative Adversarial Networks (NeurIPS 2019 · similarity 0.93)
- Testing Conditional Mean Independence Using Generative Neural Networks (ICML 2025 · similarity 0.88)
- Dual Projection Generative Adversarial Networks for Conditional Image Generation (ICCV 2021 · similarity 0.88)
- DO-GAN: A Double Oracle Framework for Generative Adversarial Networks (CVPR 2022 · similarity 0.88)
- Triple Generative Adversarial Nets (NeurIPS 2017 · similarity 0.87)
- Are GANs Created Equal? A Large-Scale Study (NeurIPS 2018 · 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.
