Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Vector Quantized Bayesian Neural Network Inference for Data Streams
Namuk Park, Taekyu Lee, Songkuk Kim
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 |
|---|---|---|---|
| Replace repeated Bayesian forward passes with exponential temporal smoothing of recent predictions | Bayesian U-Net · CamVid semantic segmentation | stream throughput (Img/Sec): 0.788 -> 9.41 | Improved |
| Replace repeated Bayesian forward passes with exponential temporal smoothing of recent predictions | Bayesian U-Net · NYUDv2 depth estimation | stream throughput (Img/Sec): 1.61 -> 13.6 | Improved |
| Apply temporal smoothing to Bayesian predictions | Bayesian U-Net · CamVid semantic segmentation | NLL: 0.276 -> 0.253 | Improved |
| Apply temporal smoothing to Bayesian predictions | Bayesian U-Net · NYUDv2 depth estimation | RMSE (m): 0.705 -> 0.700 | Improved |
| Use deterministic weights instead of a Bayesian weight distribution while retaining temporal smoothing (ablation) | VQ-BNN · CamVid semantic segmentation | NLL: 0.253±0.009 -> 0.284±0.023 | Hurt |
| Reduce the number of recent prototypes (ablation) | VQ-BNN · CamVid semantic segmentation | NLL | Hurt |
| Increase the exponential-decay timescale up to 1.25 (ablation) | VQ-BNN · CamVid semantic segmentation | NLL | Improved |
| Increase the exponential-decay timescale beyond 1.25 (ablation) | VQ-BNN · CamVid semantic segmentation | NLL | Hurt |
| Add future predictions to the temporal smoothing window (ablation) | VQ-BNN · CamVid semantic segmentation | NLL | Improved |
| Reduce video frame rate and temporal consistency (ablation) | VQ-BNN · CamVid semantic segmentation | NLL | Hurt |
| Apply temporal smoothing to deterministic predictions rather than Bayesian predictions (ablation) | DNN · CamVid semantic segmentation | NLL: 0.314 -> 0.284 | Improved |
| Apply temporal smoothing to deterministic predictions rather than Bayesian predictions (ablation) | DNN · NYUDv2 depth estimation, SegNet | RMSE (m): 0.740±0.028 -> 0.717±0.042 | Improved |
Closest papers by meaning
- Encoding the Latent Posterior of Bayesian Neural Networks for Uncertainty Quantification (TPAMI 2023 · similarity 0.90)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.89)
- Sampling-Free Learning of Bayesian Quantized Neural Networks (ICLR 2020 · similarity 0.88)
- Do Bayesian Neural Networks Actually Behave Like Bayesian Models? (ICML 2025 · similarity 0.87)
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.87)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 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.
