Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Sparse Bayesian Deep Functional Learning with Structured Region Selection
Xiaoxian Zhu, Yingmeng Li, Shuangge Ma, Mengyun Wu
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 sBayFDNN instead of linear region selectors (BFRS, SLoS, cFuSIM) on ECG | sBayFDNN · ECG | region F1: 0.412 -> 0.634 (vs SLoS) | Improved |
| use sBayFDNN instead of baselines on Tecator | sBayFDNN · Tecator | RMSE: 2.217 -> 2.138 (vs FNN) | Improved |
| use sBayFDNN instead of baselines on Bike rental | sBayFDNN · Bike rental | RMSE: 0.684 -> 0.618 (vs SLoS) | Improved |
| add structured first-layer sparsity versus plain spline-feature FNN | spline-feature DNN · ECG | RMSE: 12.991 -> 12.069 | Improved |
| increase spline bases Jn from small to moderate (80) (ablation) | sBayFDNN · Simulation (simple beta, composite g, SNR 10) | region F1: 0.5573 -> 0.9323 | Improved |
| increase spline bases Jn to very large (240) (ablation) | sBayFDNN · Simulation (simple beta, composite g, SNR 10) | region F1: 0.9323 -> 0.7583 | Hurt |
| raise PIP threshold from 0.5 to 0.9 (ablation) | sBayFDNN · Simulation macro-average | region F1: 0.8781 -> 0.4557 | Hurt |
| lower PIP threshold from 0.5 to 0.3 (ablation) | sBayFDNN · Simulation macro-average | region F1: 0.8781 -> 0.8969 | Improved |
| shrink spike variance to 1e-6 (ablation) | sBayFDNN · Simulation macro-average | region F1: 0.7982 -> 0.5463 | Hurt |
| vary prior inclusion probability lambda_n (ablation) | sBayFDNN · Simulation macro-average | region F1: 0.7868 to 0.8182 across values | No clear effect |
| use evidence-based selection of Jn | sBayFDNN · Simulation (composite g, SNR 10) | within 30% of oracle F1: 97.1% | Improved |
| increase observation grid Tobs from 200 to 2000 (ablation) | sBayFDNN · Simulation runtime (n=1000) | runtime per repeat (s): 125.1 -> 103.7 | No clear effect |
Closest papers by meaning
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · similarity 0.85)
- Deep Functional Factor Models: Forecasting High-Dimensional Functional Time Series via Bayesian Nonparametric Factorization (ICML 2024 · similarity 0.84)
- A Functional Extension of Semi-Structured Networks (NeurIPS 2024 · similarity 0.84)
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.84)
- Posterior Contraction for Sparse Neural Networks in Besov Spaces with Intrinsic Dimensionality (NeurIPS 2025 · similarity 0.84)
- Posterior Contraction Rates for sparse Kolmogorov-Arnold Networks in Anisotropic Besov Spaces (NeurIPS 2026 · 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.
