Research map / Graph learning / GNN architectures
NetGAN without GAN: From Random Walks to Low-Rank Approximations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace GAN+LSTM with low-rank approximation of logit transition matrix trained by cross-entropy (CELL) | NetGAN · CORA-ML | ROC-AUC score: 0.950 vs 0.938 | No clear effect |
| Replace GAN+sampling with CELL (low-rank logits, CE loss, no sampling) | NetGAN · CORA-ML | Training time (s): 7,478 s to 21 s | Helped |
| Use Frobenius-norm low-rank approximation of transition matrix instead of cross-entropy loss (ablation) | low-rank approximation of transition matrix · CORA-ML | ROC-AUC score: 0.709 (LR-TRANS) vs 0.950 (LR-CE) | Hurt |
| Use cross-entropy loss (KL divergence) instead of Frobenius norm for low-rank approximation (ablation) | low-rank approximation of transition matrix · CORA-ML | ROC-AUC score: 0.950 (LR-CE) vs 0.709 (LR-TRANS) | Helped |
| Remove the logit transform (LR-CE) vs keep logit space (CELL) (ablation) | CELL · CORA-ML | ROC-AUC score: 0.950 (LR-CE) vs 0.938 (CELL) | No clear effect |
| Use Frobenius low-rank approximation of adjacency matrix (LR-ADJ) instead of CE-based model (ablation) | low-rank approximation baselines · CORA-ML | ROC-AUC score: 0.561 vs 0.938 (CELL) | Hurt |
| Use cross-entropy low-rank logit model with sampling-free reconstruction in place of NetGAN | CELL graph statistics · CORA-ML | Triangle count: 1,396 (CELL) vs 1,461 (NetGAN) | No clear effect |
Closest papers by meaning
- NetGAN: Generating Graphs via Random Walks (ICML 2018 · similarity 0.96)
- GraphGAN: Graph Representation Learning With Generative Adversarial Nets (AAAI 2018 · similarity 0.90)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
- Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality (ICML 2025 · similarity 0.87)
- CG-TGAN: Conditional Generative Adversarial Networks with Graph Neural Networks for Tabular Data Synthesizing (AAAI 2025 · 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.