Research map / Reinforcement learning / Visual and representation RL
Virne: A Comprehensive Benchmark for RL-based Network Resource Allocation in NFV
Tianfu Wang, Liwei Deng, Xi Chen, Junyang Wang, Huiguo He, Zhengyu Hu, Wei Wu, Leilei Ding, Qilin Fan, Hui Xiong
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 |
|---|---|---|---|
| add action masking (ablation) | PPO-DualGAT · WX100 | RAC: 0.648 -> 0.781 | Improved |
| add action masking (ablation) | PPO-MLP · WX100 | RAC: 0.666 -> 0.719 | Improved |
| add topological features to status features (ablation) | PPO-ATT · WX100 | RAC: 0.579 -> 0.712 | Improved |
| use fixed moderate intermediate reward 0.1 instead of none (ablation) | PPO-MLP · WX100 | RAC: 0.560 -> 0.719 | Improved |
| use fixed 0.1 reward instead of adaptive normalized reward (ablation) | PPO-MLP · WX100 | RAC: 0.705 -> 0.719 | Improved |
| use PPO instead of PG or A3C for training (ablation) | PPO-DualGAT · WX100 | training return | Improved |
| use dual-GNN policy instead of MLP | PPO-DualGAT · WX100 | RAC: 71.90 -> 78.10 | Improved |
| use RL policy instead of meta-heuristic/heuristic solvers | PPO-DualGAT · WX100 | RAC / AST: LRC 0.74 vs 0.44 (MCTS); AST 0.18 vs 3.38 | Improved |
| use GCN instead of GAT on small sparse topologies | PPO-GCN · GEANT / BRAIN | RAC: GEANT 58.70 vs PPO-GAT 58.40; BRAIN 49.50 vs 44.60 | Mixed |
| use traditional heuristics instead of RL | RW-Rank etc. · WX100 | LRC | Hurt |
Closest papers by meaning
- A Deep Reinforcement Learning Perspective on Internet Congestion Control (ICML 2019 · similarity 0.86)
- SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores (ICLR 2024 · similarity 0.85)
- IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL (ICML 2026 · similarity 0.84)
- MiniHack the Planet: A Sandbox for Open-Ended Reinforcement Learning Research (NeurIPS 2021 · similarity 0.84)
- Influence-Augmented Local Simulators: a Scalable Solution for Fast Deep RL in Large Networked Systems (ICML 2022 · similarity 0.84)
- Horizon Reduction Makes RL Scalable (NeurIPS 2025 · similarity 0.83)
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.
