Logic Synthesis
Useful GNN embeddings for logic synthesis must preserve functional and Boolean structure rather than just local topology. DeepGate learns AIG gate embeddings using simulated signal probabilities and separates functional from structural rep…
1 sources - 4 claims
Useful GNN embeddings for logic synthesis must preserve functional and Boolean structure rather than just local topology. DeepGate learns AIG gate embeddings using simulated signal probabilities and separates functional from structural representations. DeepSeq2 represents sequential circuits by disentangling structural, functional, and sequential-behavior embeddings. MGVGA uses masked gate modeling and Verilog-AIG alignment because generic graph masking can break circuit equivalence.