Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Efficient Variational Inference for Sparse Deep Learning with Theoretical Guarantee
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 spike-and-slab sparse variational BNN instead of dense VBNN | Simulation I dense teacher · Simulation I dense teacher | Sparsity(%): 100 ± 0.00 -> 6.45 ± 0.83 | Helped |
| Use spike-and-slab sparse variational BNN instead of dense VBNN | Simulation I dense teacher · Simulation I dense teacher | Test RMSE: 1.00 ± 0.00 -> 1.01 ± 0.00 | No clear effect |
| Use spike-and-slab sparse variational BNN instead of AGP | Simulation I sparse teacher · Simulation I sparse teacher | Test RMSE: 1.02 ± 0.00 -> 1.00 ± 0.01 | Helped |
| Use spike-and-slab sparse variational BNN instead of VBNN | Simulation I sparse teacher · Simulation I sparse teacher | FPR(%): 100 ± 0.00 -> 0.00 ± 0.00 | Helped |
| Use spike-and-slab sparse variational BNN instead of VBNN | Simulation I sparse teacher · Simulation I sparse teacher | 95% Coverage (%): 90.7 ± 8.15 -> 96.4 ± 4.73 | Helped |
| Use spike-and-slab sparse variational BNN instead of VD | Simulation II sparse nonlinear function · Simulation II sparse nonlinear function | Test RMSE: 1.43 ± 0.19 -> 1.21 ± 0.05 | Helped |
| Use spike-and-slab sparse variational BNN instead of VBNN | Simulation II sparse nonlinear function · Simulation II sparse nonlinear function | Test RMSE: 1.99 ± 0.49 -> 1.21 ± 0.05 | Helped |
| Use spike-and-slab sparse variational BNN instead of VD | Simulation II sparse nonlinear function · Simulation II sparse nonlinear function | FPR(%): 98.6 ± 1.22 -> 0.00 ± 0.21 | Helped |
| Increase the prior inclusion probability λ (ablation) | SVBNN · Simulation I sparse teacher | Training RMSE | Helped |
| Increase λ away from the suggested λopt (ablation) | SVBNN · Simulation I sparse teacher | Test RMSE | Mixed |
| Decrease λ below the suggested λopt (ablation) | SVBNN · Simulation I sparse teacher | Test RMSE | Hurt |
| Replace inverse-CDF reparameterization with Gumbel-softmax approximation (ablation) | SVBNN · Simple model comparison | Results | No clear effect |
Closest papers by meaning
- Convergence Rates of Variational Inference in Sparse Deep Learning (ICML 2020 · similarity 0.94)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · similarity 0.90)
- Bayesian Online Natural Gradient (BONG) (NeurIPS 2024 · similarity 0.88)
- VIKING: Deep variational inference with stochastic projections (NeurIPS 2025 · similarity 0.88)
- Variational Bayesian Pseudo-Coreset (ICLR 2025 · similarity 0.88)
- Sparse Inducing Points in Deep Gaussian Processes: Enhancing Modeling with Denoising Diffusion Variational Inference (ICML 2024 · similarity 0.88)
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.