Integrated circuit yield optimization plays a vital role in ensuring reliable semiconductor manufacturing, directly impacting both product quality and production costs. Current approaches to yield optimization face two fundamental challenges that limit their practical effectiveness. First, yield estimation requires intensive computational resources. Second, traditional black-box optimization methods inefficiently allocate these resources across design candidates. Most existing approaches compound these issues by performing detailed yield estimations uniformly across all candidates, regardless of their potential quality. To address these limitations, we introduce a novel precision-aware yield optimization framework that intelligently adapts computational resource allocation based on each design candidate’s predicted performance. Our approach moves beyond simple simulation counting by incorporating a Figure of Merit (FoM) as a continuous quality metric. By combining a Continuous AutoRegression model to characterize the relationship between true yield and precision levels with a sophisticated multi-fidelity acquisition strategy, our framework achieves optimal resource distribution. Experimental validation on four industry-standard benchmark circuits demonstrates that our method converges with fewer than 1,000 simulations, reducing simulation costs by over $10 \times$ while achieving better final designs and robustness than state-of-the-art high-fidelity approaches.

Federated learning has been introduced into network intrusion detection to address the data silos among intrusion detection systems in distributed environments. However, existing federated learning-based methods suffer from the problem of class-imbalanced traffic. Inspired by the Bayes theorem, we propose an adaptive cost weight-enabled federated distillation method for distributed network intrusion detection. The proposed method comprises a federated distillation module and a cost-sensitive module. The federated distillation module combines knowledge distillation and federated learning to establish a distributed intrusion detection system, ensuring data privacy while reducing the interaction overhead. The cost-sensitive module optimizes the federated distillation module by calculating posterior probabilities using prior probabilities and real-time observations from the client training process. By adjusting the model’s attention to different samples based on the posterior probabilities, the method enhances the detection accuracy of intrusion detection models for classes with few samples. Experimental results show that the proposed method achieves state-of-the-art performance on the NSL-KDD dataset, with accuracy improvements of 5.6% and 3.6% compared to the FedAvg and FedProx algorithms, respectively.