Abstract
A machine learning-based hybrid system can utilize multi-scale spatial-temporal power grid attention for real-time dynamic ohmic (IR)-drop prediction. The system employs a two-stage hierarchical architecture combining convolutional neural networks to capture global spatial hotspot contexts and gradient-boosted decision trees to predict instance-specific magnitudes. By shifting from a traditional batch verification process to a real-time predictive approach, the system significantly accelerates design convergence while maintaining high accuracy.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Picozzi, Google Case Managers; Gotra, Vishant; and Pendyala, Prateek, "Machine Learning-Based Hybrid System with Multi-Scale Spatial-Temporal Power Grid Attention for Real-Time Dynamic IR-Drop Prediction", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9983