Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Discriminative Bayesian Networks from High-Dimensional Continuous Neuroimaging Data
Luping Zhou, Lei Wang, Lingqiao Liu, Philip Ogunbona, Dinggang G. Shen
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 |
|---|---|---|---|
| optimize all Theta together with the proposed ordering-based DAG constraint (OR-SGBN WHOLE) instead of column-wise BCD | H-SGBN · feature-order permutation, Theta1 | distance to original Theta: 1.91 -> 0.08 | Improved |
| optimize all Theta together with the proposed ordering-based DAG constraint instead of column-wise BCD | H-SGBN · feature-order permutation, Theta1 | correlation R: 0.6828 -> 0.9996 | Improved |
| optimize all Theta together with the proposed ordering-based DAG constraint instead of column-wise BCD | H-SGBN · feature-order permutation, Theta2 | correlation R: 0.6396 -> 0.9981 | Improved |
| use the old DAG constraint while optimizing all Theta together (H-SGBN WHOLE) | H-SGBN · ADNI neuroimaging | convergence to a DAG | Hurt |
| use OR-SGBN instead of L1MB for structure learning | L1MB · Carpo benchmark | total edge errors: 125.74 -> 25.58 | Improved |
| use OR-SGBN instead of L1MB for structure learning | L1MB · Factors benchmark | total edge errors: 101.48 -> 54.82 | Improved |
| use OR-SGBN instead of L1MB for structure learning | L1MB · Chain benchmark | total edge errors: 5.32 -> 7.04 | Hurt |
| use OR-SGBN instead of L1MB for structure learning | L1MB · Water benchmark | total edge errors: 92.86 -> 93.08 | No clear effect |
| use OR-SGBN instead of L1MB for structure learning | L1MB · Hailfinder benchmark | falsely identified edges: 60.00 -> 28.66 | Improved |
| add discriminative learning (KL-SGBN, MM-SGBN) to individually learned SGBNs | OR-SGBN / H-SGBN · MRI, PET, MRI-II | test accuracy: 10% ~ 20% | Improved |
| use MM-SGBN instead of KL-SGBN with SGBN classifiers | KL-SGBN · MRI, PET, MRI-II | test accuracy | Improved |
| use KL-SGBN instead of MM-SGBN features with SVM classifiers | MM-SGBN · MRI, PET, MRI-II | SVM test accuracy | Improved |
Closest papers by meaning
- Understanding Non-linearity in Graph Neural Networks from the Bayesian-Inference Perspective (NeurIPS 2022 · similarity 0.86)
- Encoding the Latent Posterior of Bayesian Neural Networks for Uncertainty Quantification (TPAMI 2023 · similarity 0.86)
- Bayesian Scalar-on-Image Regression with a Spatially Varying Single-layer Neural Network Prior (JMLR 2025 · similarity 0.86)
- Testing for Differences in Gaussian Graphical Models: Applications to Brain Connectivity (NeurIPS 2016 · similarity 0.86)
- Neural Network Parameter-optimization of Gaussian Pre-marginalized Directed Acyclic Graphs (JMLR 2026 · similarity 0.86)
- An Asymptotically Optimal Coordinate Descent Algorithm for Learning Bayesian Networks from Gaussian Models (JMLR 2025 · similarity 0.85)
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.
