PREDICTION OF ELECTRONIC CONNECTIVITY BASED ON GRAPH ENCODER MODELS

Intelligently assisted analog integrated circuit design requires a large amount of circuit netlist annotation work. In this paper, we propose a netlist annotation method based on graph attention coding models. The method treated component ports as nodes in graph data, and established and trained the graph neural network SGL-WalkPool, which could quickly annotate the connection relationships between electronic components from images containing analog circuit schematics. In addition, we proposed the bypass structure SLG and the S-mish activation function to improve the graph encoder model. Experimental results show that the improved algorithm proposed in this paper achieves better performance than comparison algorithms on both custom dataset and public dataset.