Research map / Probabilistic and scientific modeling / Variational and generative sampling
Poisson-Randomised DirBN: Large Mutation is Needed in Dirichlet Belief Networks
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 |
|---|---|---|---|
| Add component-wise Poisson counts D (Pois-DirBN-D) to the layer-wise connection (ablation) | DirBN (SDREM) · Citeer / PPI | AUC: Citeer 77.9 -> 80.5; PPI 89.2 -> 92.4 | Helped |
| Add only node-wise Poisson counts C (Pois-DirBN-C) (ablation) | DirBN (SDREM) · Citeer / PPI | AUC: Citeer 77.9 -> 78.3; PPI 89.2 -> 88.7 | No clear effect |
| Use both C and D (Pois-DirBN-CD) instead of DirBN | DirBN (SDREM) · Citeer, Cora, Pubmed, PPI | AUC: Citeer 77.9 -> 82.9; Cora 83.2 -> 89.3; Pubmed 84.5 -> 88.4; PPI 89.2 -> 92.3 | Helped |
| Replace the integer-marginalised formulation (Integer-DirBN) vs Pois-DirBN-CD (ablation) | Pois-DirBN · PPI | AUC: PPI 92.3 -> 89.2 | Hurt |
| Increase number of layers from 2 to 3 or more in Pois-DirBN-CD (ablation) | Pois-DirBN-CD · Citeer, Cora, Pubmed, PPI (Figure 3) | AUC: L=2 significantly worse than other cases; L=10 best | Helped |
| Increase number of communities K from 5 up to 20 in Pois-DirBN-CD (ablation) | Pois-DirBN-CD · Citeer, Cora, Pubmed, PPI (Figure 3) | AUC and average precision: increase not significant when K > 20 | Helped |
Closest papers by meaning
- Multimodal Poisson Gamma Belief Network (AAAI 2018 · similarity 0.87)
- Scalable Deep Generative Relational Model with High-Order Node Dependence (NeurIPS 2019 · similarity 0.87)
- Deep Latent Dirichlet Allocation with Topic-Layer-Adaptive Stochastic Gradient Riemannian MCMC (ICML 2017 · similarity 0.86)
- Convolutional Poisson Gamma Belief Network (ICML 2019 · similarity 0.85)
- Compositional Discrete Latent Code for High Fidelity, Productive Diffusion Models (NeurIPS 2025 · similarity 0.82)
- Discrete Distribution Networks (ICLR 2025 · similarity 0.82)
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.