基于稀疏增强KAN网络的模拟电路多目标优化方法及系统
By constructing a surrogate model using a sparse augmented KAN network, the problems of insufficient efficiency and solution set quality in multi-objective optimization of analog circuits are solved, achieving efficient and interpretable multi-objective optimization and improving the convergence and distribution uniformity of the Pareto optimal solution set.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QINGDAO UNIV OF SCI & TECH
- Filing Date
- 2026-05-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multi-objective optimization techniques for analog circuits cannot simultaneously balance optimization efficiency, solution quality, and solution interpretability. Traditional neural networks lack fitting accuracy and generalization ability, evolutionary algorithms have insufficiently differentiated search strategies, and surrogate models are prone to local accuracy degradation, making it difficult to adapt to the engineering optimization needs of complex circuits.
A sparse augmented KAN network is used to construct a proxy model. Through sparse gating mechanism and multi-output head architecture, design parameter sensitivity information is extracted. Combined with adaptive variable asynchronous length and hierarchical incremental learning mechanism, differentiated search and model update are achieved, improving fitting accuracy and search efficiency.
It significantly improves the efficiency and solution set quality of multi-objective optimization of analog circuits, enhances the convergence and uniformity of Pareto optimal solution sets, reduces computational overhead, and provides interpretable quantization guidance.
Smart Images
Figure CN122197785B_ABST