Inverse Learning-Rate Scaling
Cross-source consensus on Inverse Learning-Rate Scaling from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Benefits
Risks & contraindications
Evidence quality
Highlighted claims
- FedQueue scales each client's learning rate inversely with its local step budget to equalize effective local displacement. — FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
- Inverse learning-rate scaling addresses instability caused by different local step counts across clients. — FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
- Unscaled local SGD could let clients with larger local step counts move farther from the broadcast model and dominate aggregation. — FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
- Removing inverse learning-rate scaling lowered final accuracy and slightly increased time-to-target in ablation experiments. — FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training
- Scaling the learning rate inversely with the local step count supports the convergence proof by bounding local displacement. — FedQueue: Queue-Aware Federated Learning for Cross-Facility HPC Training