Research map / Probabilistic and scientific modeling / Variational and generative sampling
The Variational Bandwidth Bottleneck: Stochastic Evaluation on an Information Budget
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 VBB instead of Emergent Communication | multiagent communication · 6 agents | distance to destination: 4.85 -> 4.72 | Helped |
| Use VBB instead of Emergent Communication | multiagent communication · 10 agents | distance to destination: 5.44 -> 5.22 | Helped |
| Use VBB instead of InfoBot | multiagent communication · 6 agents | distance to destination: 4.81 -> 4.72 | Helped |
| Use VBB instead of InfoBot | multiagent communication · 10 agents | distance to destination: 5.32 -> 5.22 | Helped |
| Use VBB rather than RAM with 6 locations | recurrent visual attention · MNIST | classification error: 1.55% -> 1.42% | Helped |
| Use VBB rather than RAM with 6 locations | recurrent visual attention · 60 * 60 Cluttered MNIST | classification error: 4.3% -> 3.8% | Helped |
| Use VBB rather than VIB | recurrent visual attention · MNIST | classification error: 1.58% -> 1.42% | Helped |
| Use VBB rather than VIB | recurrent visual attention · 60 * 60 Cluttered MNIST | classification error: 4.2% -> 3.8% | Helped |
| Use VBB to determine privileged-input access | Navigation Env · Navigation Env | average information transmission: 4.45 (100%) -> 3.92 (20%) | Helped |
| Use VBB to determine privileged-input access | Sequential MNIST · Sequential MNIST | average information transmission: 3.56 (100%) -> 3.22 (46%) | Helped |
| Use VBB to determine privileged-input access | Model Based RL · Model Based RL | average information transmission: 7.12 (100%) -> 6.94 (15%) | Helped |
| Learn the Bernoulli access parameter with REINFORCE instead of the proposed method (ablation) | privileged-input access · Navigation Env | average information transmission: 3.92 (20%) -> 5.34 (74%) | Hurt |
Closest papers by meaning
- InfoBot: Transfer and Exploration via the Information Bottleneck (ICLR 2019 · similarity 0.89)
- Robust Predictable Control (NeurIPS 2021 · similarity 0.89)
- Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneck (NeurIPS 2019 · similarity 0.88)
- Information-Directed Pessimism for Offline Reinforcement Learning (ICML 2024 · similarity 0.86)
- Is Value Learning Really the Main Bottleneck in Offline RL? (NeurIPS 2024 · similarity 0.86)
- Mitigating Partial Observability in Sequential Decision Processes via the Lambda Discrepancy (NeurIPS 2024 · 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.