Research map / Probabilistic and scientific modeling / Dynamical systems
Extrapolative Continuous-time Bayesian Neural Network for Fast Training-free Test-time Adaptation
Hengguan Huang, Xiangming Gu, Hao Wang, Chang Xiao, Hongfu Liu, Ye Wang
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 |
|---|---|---|---|
| remove upper bound of temporal-domain-invariant loss (eli) (ablation) | ECBNN · Streaming Rotating MNIST->USPS target | accuracy: 53.6 -> 60.9 | Improved |
| add temporal-domain-invariant loss upper bound to ECBNN (ablation) | ECBNN · Streaming Rotating MNIST->USPS out-of-domain | accuracy: 29.9 -> 38.5 | Improved |
| add PFDE loss (lde) on top of ELBO (ablation) | ECBNN · Streaming Rotating MNIST->USPS target | accuracy: 43.2 -> 53.6 | Improved |
| add PFDE loss (lde) on top of ELBO (ablation) | ECBNN · Streaming Rotating MNIST->USPS out-of-domain | accuracy: 28.0 -> 29.9 | Improved |
| use ECBNN instead of ADDA | ADDA · Streaming Rotating MNIST->USPS target | accuracy: 52.0 -> 60.9 | Improved |
| use ECBNN instead of ADDA | ADDA · Streaming Rotating MNIST->USPS out-of-domain | accuracy: 34.3 -> 38.5 | Improved |
| use ECBNN instead of Source-Only | Source-Only · Streaming Rotating MNIST->USPS target | accuracy: 28.8 -> 60.9 | Improved |
| use ECBNN instead of EDA online adaptation | EDA · OuluVS2 target | accuracy: 70.3 -> 75.3 | Improved |
| use ECBNN instead of EDA online adaptation | EDA · OuluVS2 | run time per batch (s): 0.23 -> 0.01 | Improved |
| use ECBNN instead of DANN | DANN · OuluVS2 target | accuracy: 71.5 -> 75.3 | Improved |
| use ECBNN instead of Source-Only on source domain | Source-Only · Streaming Rotating MNIST->USPS source | accuracy: 97.8 -> 97.1 | No clear effect |
| increase observed history length at test time | ECBNN · Growing Circle | accuracy | Improved |
Closest papers by meaning
- Test-time Adaptation in Non-stationary Environments via Adaptive Representation Alignment (NeurIPS 2024 · similarity 0.84)
- ViDA: Homeostatic Visual Domain Adapter for Continual Test Time Adaptation (ICLR 2024 · similarity 0.84)
- Efficient Source-Free Time-Series Adaptation via Parameter Subspace Disentanglement (ICLR 2025 · similarity 0.84)
- Backpropagation-free Network for 3D Test-time Adaptation (CVPR 2024 · similarity 0.84)
- Bayesian Neural Scaling Law Extrapolation with Prior-Data Fitted Networks (ICML 2025 · similarity 0.84)
- Non-Stationary Learning of Neural Networks with Automatic Soft Parameter Reset (NeurIPS 2024 · similarity 0.84)
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.
