Research map / Architectures and efficiency / RNNs and normalization
Parsimonious Bayesian deep networks
Mingyuan Zhou
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 PBDN-AICε with Gibbs inference rather than RBF SVM | binary classification benchmarks · image | classification error: 2.84 -> 2.30 | Improved |
| Use PBDN-AICε with Gibbs inference rather than RBF SVM | binary classification benchmarks · banana | classification error: 10.85 -> 10.97 | No clear effect |
| Use PBDN-AICε with Gibbs inference rather than RBF SVM | binary classification benchmarks · ijcnn1 | classification error: 4.01 -> 4.06 | Hurt |
| Use AICε-Gibbs rather than AIC-Gibbs for depth selection (ablation) | PBDN · image | classification error: 2.36 -> 2.30 | Improved |
| Use AICε-Gibbs rather than AIC-Gibbs for depth selection (ablation) | PBDN · ijcnn1 | classification error: 4.33 -> 4.06 | Improved |
| Use AICε-Gibbs rather than AIC-Gibbs for depth selection (ablation) | PBDN · waveform | classification error: 11.42 -> 11.45 | No clear effect |
| Use four hidden layers rather than one (ablation) | PBDN · banana | classification error: 22.54 -> 10.85 | Improved |
| Use two hidden layers rather than one (ablation) | PBDN · image | classification error: 3.32 -> 2.43 | Improved |
| Increase PBDN depth from one to two layers (ablation) | PBDN · breast cancer | classification error: 29.22 -> 30.26 | Hurt |
| Train a jointly optimized DNN using the architecture inferred by PBDN instead of greedy layer-wise PBDN (ablation) | PBDN-inferred architectures · eight benchmark datasets | mean SVM-normalized error: 1.047, 1.011, 1.076, and 1.144 | Mixed |
| Stack CPM layers rather than use PBDN's iSHM layers (ablation) | CPM · waveform | classification error: 0.131, 0.177, 0.223 | Hurt |
| Stack CPM layers rather than use PBDN's iSHM layers (ablation) | CPM · image | classification error: 0.046, 0.080, 0.216 | Hurt |
Closest papers by meaning
- Nonparametric Bayesian Deep Networks with Local Competition (ICML 2019 · similarity 0.89)
- Bayesian Deep Learning via Subnetwork Inference (ICML 2021 · similarity 0.86)
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.86)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · similarity 0.86)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · similarity 0.86)
- Partially Stochastic Infinitely Deep Bayesian Neural Networks (ICML 2024 · 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.
