Research map / Architectures and efficiency / Mixture of experts
FT-MoE: Sustainable-learning Mixture of Experts for Fault-Tolerant Computing
Wenjing Xiao, Wenhao Song, Miaojiang Chen, Min Chen
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 EAGate adaptive expert selection instead of static gating (ablation) | FT-MoE · Fault classification training accuracy | best accuracy (%): 87.956% -> 90.479% | Improved |
| add online tuning (expert add/remove) during deployment (ablation) | FT-MoE · Online running, fault detection | accuracy (%): without tuning drops to 52.624% | Improved |
| add online tuning (expert add/remove) during deployment (ablation) | FT-MoE · Online running, fault classification | accuracy (%): without tuning drops to 57.223% | Improved |
| add cross multi-head attention (CMHA) fusing MoE and graph paths (ablation) | FT-MoE · Fault classification t-SNE | cluster separation (qualitative) | Improved |
| use FT-MoE instead of PreGAN+ for fault classification | fault prediction · Fault classification (new edge dataset) | HR: 0.6250 -> 0.6496 | Improved |
| use FT-MoE instead of PreGAN+ for fault detection accuracy | fault prediction · Fault detection (new edge dataset) | accuracy: 0.9060 -> 0.9053 | No clear effect |
| use FT-MoE instead of PreGAN+ for detection recall | fault prediction · Fault detection (new edge dataset) | recall: 0.9180 -> 0.9322 | Improved |
| use FT-MoE fault predictions to drive GAN scheduling instead of PreGAN+ | scheduling · Scheduling QoS (COSCO simulation) | average energy (KW-hr): 16.4324 -> 14.1641 | Improved |
Closest papers by meaning
- A Theory of Fault-Tolerant Learning (ICML 2024 · similarity 0.82)
- Certain Head, Uncertain Tail: Expert-Sample for Test-Time Scaling in Fine-Grained MoE (ICML 2026 · similarity 0.81)
- Typicalness-Aware Learning for Failure Detection (NeurIPS 2024 · similarity 0.80)
- Efficient Deweahter Mixture-of-Experts with Uncertainty-Aware Feature-Wise Linear Modulation (AAAI 2024 · similarity 0.80)
- A Provably Effective Method for Pruning Experts in Fine-tuned Sparse Mixture-of-Experts (ICML 2024 · similarity 0.80)
- Joint MoE Scaling Laws: Mixture of Experts Can Be Memory Efficient (ICML 2025 · similarity 0.79)
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.
