Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Continual Learning via Sequential Function-Space Variational Inference
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 sequential function-space variational inference instead of FROMP | Continual-learning neural networks · Split MNIST (MH) | accuracy: 99.00% ± 0.04 -> 99.54% ± 0.04 | Helped |
| use sequential function-space variational inference instead of FROMP | Continual-learning neural networks · Split Fashion MNIST (MH) | accuracy: 99.00% ± 0.03 -> 99.19% ± 0.02 | Helped |
| use a larger neural network with S-FSVI (ablation) | S-FSVI · Permuted MNIST (SH) | accuracy: 95.76% ± 0.02 -> 97.50% ± 0.01 | Helped |
| use a larger neural network with S-FSVI (ablation) | S-FSVI · Split Fashion MNIST (MH) | accuracy: 99.19% ± 0.02 -> 99.16% ± 0.03 | No clear effect |
| remove the coreset (ablation) | S-FSVI · Split MNIST (SH) | accuracy: 92.87% ± 0.14 -> 20.15% ± 0.52 | Hurt |
| remove the coreset (ablation) | S-FSVI · Split MNIST (MH) | accuracy: 99.54% ± 0.04 -> 99.62% ± 0.02 | No clear effect |
| use one sample per class as a coreset (ablation) | S-FSVI · Permuted MNIST (SH) | accuracy: 95.76% ± 0.02 -> 89.59% ± 0.30 | Hurt |
| use sequential function-space variational inference instead of FROMP | Continual-learning neural networks · Split CIFAR | test accuracy: 76.2% ± 0.2 -> 77.6% ± 0.2 | Helped |
| use sequential function-space variational inference instead of FRCL | Continual-learning neural networks · Sequential Omniglot | test accuracy: 81.47% ± 1.6 -> 83.29% ± 1.2 | Helped |
| use sequential function-space inference instead of parameter-space variational inference | S-FSVI versus VCL · Split MNIST and permuted MNIST | accuracy | Helped |
| use no coreset (ablation) | S-FSVI · Permuted MNIST (SH) | accuracy: 95.76% ± 0.02 -> 84.06% ± 0.46 | Hurt |
Closest papers by meaning
- Function-space Parameterization of Neural Networks for Sequential Learning (ICLR 2024 · similarity 0.89)
- A Statistical Theory of Regularization-Based Continual Learning (ICML 2024 · similarity 0.89)
- Temporal-Difference Variational Continual Learning (NeurIPS 2025 · similarity 0.88)
- Optimal Rates in Continual Linear Regression via Increasing Regularization (NeurIPS 2025 · similarity 0.88)
- Globally Convergent Variational Inference (NeurIPS 2024 · similarity 0.88)
- Learning Continually by Spectral Regularization (ICLR 2025 · 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.