Research map / Probabilistic and scientific modeling / Dynamical systems
Learning and Policy Search in Stochastic Dynamical Systems with Bayesian Neural Networks
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 |
|---|---|---|---|
| Train the dynamics BNN with α=0.5 rather than VB | BNN policy search · WetChicken | average reward: -2.67 -> -2.37 | Helped |
| Train the dynamics BNN with α=0.5 rather than VB | BNN policy search · Turbine | average reward: -0.45 -> -0.41 | Helped |
| Train the dynamics BNN with α=0.5 rather than α=1.0 | BNN policy search · Industrial | average reward: -171.1 -> -174.2 | Hurt |
| Use α=0.5 rather than VB | Bayesian neural network dynamics model · WetChicken | log-likelihood: -1.140 -> -1.057 | Helped |
| Use α=0.5 rather than VB | Bayesian neural network dynamics model · Turbine | log-likelihood: -0.775 -> -0.746 | Helped |
| Use α=0.5 rather than α=1.0 | Bayesian neural network dynamics model · Industrial | log-likelihood: 1.326 -> 1.328 | No clear effect |
| Keep q(z) fixed to the prior instead of learning q(z) (ablation) | BNN trained with α=0.5 · Bi-modal prediction problem | log-likelihood: -3.05 -> -2.10 | Helped |
| Keep q(z) fixed to the prior instead of learning q(z) (ablation) | BNN trained with α=0.5 · Heteroskedastic prediction problem | log-likelihood: -2.05 -> -1.78 | Helped |
| Use α=0.5 rather than VB while q(z) is fixed to the prior (ablation) | Bayesian neural network · Bi-modal prediction problem | log-likelihood: -3.05 -> -2.10 | Helped |
| Use α=0.5 rather than VB while q(z) is fixed to the prior (ablation) | Bayesian neural network · Heteroskedastic prediction problem | log-likelihood: -2.05 -> -1.78 | Helped |
Closest papers by meaning
- Local policy search with Bayesian optimization (NeurIPS 2021 · similarity 0.89)
- Bayesian Exploration Networks (ICML 2024 · similarity 0.88)
- Learning Optimal Deterministic Policies with Stochastic Policy Gradients (ICML 2024 · similarity 0.88)
- Reinforcement Learning with Imperfect Transition Predictions: A Bellman-Jensen Approach (NeurIPS 2025 · similarity 0.88)
- Augmented Bayesian Policy Search (ICLR 2024 · similarity 0.88)
- Stochastic Q-learning for Large Discrete Action Spaces (ICML 2024 · similarity 0.87)
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.