DRV Prediction Using Quantum-Classical Hybrid Convolutional Neural Network
Xu Zhang, Haochang Tian, Chengkai Wang, Weiqing Ji, Mingyang Kou, Hailong Yao · 2024
In very large-scale integration (VLSI) design, accurately predicting design rule violation (DRV) before the routing stage has been a significant challenge. Traditional machine learning-based methods have shown promise but are often computationally intensive. This work first introduces the concept of incorporating quantum computing into DRV prediction through the development of a novel quantum-classical hybrid convolutional neural network, named QuDRV. The proposed model leverages the unique properties of quantum mechanics to perform convolution operations, which are particularly well-suited for handling the complex placement and routing correlations found in chip designs. Experimental results demonstrate that QuDRV achieves similar prediction accuracy while maintaining a simple network structure compared with classical counterparts. Although current quantum hardware limitations result in longer runtime, future advancements in quantum technology are expected to significantly improve both the accuracy and speed of QuDRV.