This paper proposes AIGer, a novel model utilizing And-Inverter Graphs (AIGs) to enhance the efficiency of logic circuit design automation in the field of electronic design automation (EDA). To address the shortcomings of existing models due to the complex structure and large number of nodes of AIGs, which simultaneously model functional and structural characteristics and lack dynamic information propagation capabilities, AIGer consists of a node logic feature initialization embedding component and an AIGs feature learning network component. The former enables efficient node embedding by projecting logical nodes such as AND and NOT into an independent semantic space, while the latter utilizes a heterogeneous graph convolutional network to better represent the original structure and information of AIGs. It designs a dynamic relational weight matrix and a differentiated information aggregation method. Experimental results demonstrate that AIGer significantly improves the MAE and MSE compared to the best-performing existing models in signal probability prediction (SSP) and truth table distance prediction (TTDP) tasks.
