Federated learning framework enhances lithography hotspot detection with hybrid knowledge distillation

Researchers have introduced FedKD-hybrid, a new framework designed to improve federated learning for lithography hotspot detection. This approach combines parameter aggregation with knowledge distillation, using a public dataset to enable clients to exchange both model parameters and logits. Experiments on benchmark and real-world datasets show that FedKD-hybrid surpasses existing methods in effectiveness and robustness. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Enhances privacy-preserving collaborative learning for specialized industrial applications like lithography hotspot detection.

RANK_REASON This is a research paper introducing a new framework for federated learning.

Read on arXiv cs.LG →

• FedKD-hybrid
• ICCAD-2012
• Yanli Li

• paper
• other