Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Specifying Weight Priors in Bayesian Deep Neural Networks with Empirical Bayes
Ranganath Krishnan, Mahesh Subedar, Omesh Tickoo
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 |
|---|---|---|---|
| Set Gaussian weight-prior means and variational initialization from pretrained deterministic DNN weights | MFVI with MOPED · UCF-101 | validation accuracy: 0.029 -> 0.867 | Improved |
| Set Gaussian weight-prior means and variational initialization from pretrained deterministic DNN weights | MFVI with MOPED · UrbanSound8K | validation accuracy: 0.143 -> 0.819 | Improved |
| Set Gaussian weight-prior means and variational initialization from pretrained deterministic DNN weights | MFVI with MOPED · CIFAR-10 | validation accuracy: 0.896 -> 0.927 (ResNet-56); 0.878 -> 0.916 (ResNet-20) | Improved |
| Use MOPED weight priors instead of random-initialized priors | MFVI · UCF-101 | AUPR: 0.0174 -> 0.9186 | Improved |
| Use MOPED weight priors instead of random-initialized priors | MFVI · UrbanSound8K | AUROC: 0.551 -> 0.9811 | Improved |
| Use MOPED weight priors instead of random-initialized priors | MFVI · MNIST | AUROC: 0.9999 -> 0.9999 | No clear effect |
| Use MOPED weight priors instead of random-initialized priors | MFVI · Diabetic Retinopathy | accuracy at 50% data retained: 0.881 -> 0.937 | Improved |
| Use MOPED weight priors instead of random-initialized priors | MFVI · Diabetic Retinopathy | AUC at 100% data retained: 0.821 -> 0.883 | Improved |
| Use MOPED weight priors instead of random-initialized priors | MFVI · Diabetic Retinopathy | accuracy at 100% data retained: 0.843 -> 0.857 | Improved |
| Initialize approximate-posterior variance using a weight-scaled perturbation factor δ (ablation) | MOPED · CIFAR-10 and UCF-101 | AUPR | Mixed |
Closest papers by meaning
- Posterior and Variational Inference for Deep Neural Networks with Heavy-Tailed Weights (JMLR 2025 · similarity 0.90)
- The Deep Weight Prior (ICLR 2019 · similarity 0.88)
- Liberty or Depth: Deep Bayesian Neural Nets Do Not Need Complex Weight Posterior Approximations (NeurIPS 2020 · similarity 0.88)
- Deterministic Variational Inference for Robust Bayesian Neural Networks (ICLR 2019 · similarity 0.88)
- Variational Bayesian Pseudo-Coreset (ICLR 2025 · similarity 0.88)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · similarity 0.86)
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.
