Research map / Continual learning / Online and replay-based
BAM: Bayes with Adaptive Memory
Josue Nassar, Jennifer Rogers Brennan, Ben Evans, Kendall Lowrey
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 BAM to select relevant training domains for adaptation | Multivariate linear regression · Rotated MNIST test domains | accuracy: 55% ± 3.7% -> 71.8% ± 5.2% | Improved |
| Use parallel selection instead of bottom-up greedy selection (ablation) | BAM · Rotated MNIST | accuracy | Hurt |
| Add regularization (ablation) | BAM · Non-stationary parameter inference | smoothness of posterior mean estimates | Improved |
| Add regularization (ablation) | BAM · Non-stationary parameter inference | posterior confidence | Hurt |
| Use BAM memory to recall previously encountered environments | BAM · Episodic Cartpole control | expected reward | Improved |
| Use BAM memory rather than resetting to the base prior after a detected change | BAM + BOCD · Continual Cartpole control | expected reward | Improved |
| Combine UCB exploration with BAM adaptation | Non-stationary bandit algorithms · Non-stationary bandits, 10 and 50 arms | cumulative regret | Improved |
| Use BOCD with Thompson sampling | Non-stationary bandit algorithms · Non-stationary bandits, 50 arms | cumulative regret | Hurt |
| Use unregularized BAM instead of regularized BAM (ablation) | BAM · Non-stationary parameter inference | smoothness of posterior mean estimates | Hurt |
Closest papers by meaning
- Learning in Markov Games with Adaptive Adversaries: Policy Regret, Fundamental Barriers, and Efficient Algorithms (NeurIPS 2024 · similarity 0.87)
- Efficient Adaptive Data Acquisition via Pretrained Belief Representations (NeurIPS 2026 · similarity 0.86)
- Online Bernstein-von Mises theorem (NeurIPS 2026 · similarity 0.86)
- ForecastCompass: Guiding Agentic Forecasting with Adaptive Factor Memory (NeurIPS 2026 · similarity 0.86)
- The Wasserstein Believer: Learning Belief Updates for Partially Observable Environments through Reliable Latent Space Models (ICLR 2024 · similarity 0.85)
- Temporal-Difference Variational Continual Learning (NeurIPS 2025 · 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.
