DiffPattern: Layout Pattern Generation via Discrete Diffusion
DescriptionDeep generative models dominate the existing literature in layout pattern generation. However, leaving the guarantee of legality to an inexplicable neural network is not a sound choice. In this paper, we propose DiffPattern to generate reliable layout patterns with three new strategies: 1) A lossless layout pattern representation method for efficient computation. 2) A novel diverse deep topology generation method via discrete diffusion method. 3) A white-box pattern assessment method to generate legal patterns given desired design rules. Experiments on the standard benchmark show that DiffPattern significantly outperforms existing baselines and is able to synthesize reliable layout patterns.
Event Type
Research Manuscript
TimeThursday, July 13th10:55am - 11:10am PDT
Location3002, 3rd Floor
Design for Manufacturing and Reliability