Alisa Stiballe will give the initial talk on her master's thesis in cooperation with Siemens EDA on Wednesday, 19th of November at 10.00am. The title of the master thesis is:
"An ML approach to optimize simulations in defect-oriented test".
The talk will be held in English. Everybody interested is invited to participate in the talk:
Abstract: Standard cell characterization in defect‑oriented test is highly time‑consuming and becomes infeasible for modern large‑input standard cells. The main bottleneck is the extensive set of analog simulations, including every pattern–defect combination to construct the Defect Detection Matrix (DDM). This thesis presents a framework that accelerates the characterization process by applying a Graph Neural Network (GNN). The GNN operates directly on the graph‑structured transistor netlist and predicts the DDM based on switch-level and limited analog simulation data, thereby reducing the overall cell characterization efforts while preserving a high defect coverage.