Presentation
Machine learning Assisted Design Rule Debug and Rule Ranking Automation
TimeWednesday, December 8th5:00pm - 6:00pm PST
LocationLevel 2 - Exhibit Hall
Event Type
Designer, IP and Embedded Systems Track Poster Networking Reception
Presented In-Person
Presenter(s)
DescriptionDesign Technology Co-optimization (DTCO) is the key driver for advanced node process development. Design rules change request design exploration are evaluated by the process development lithography team. This paper developed a Machine learning assisted design rule debug system to help technologist at the design rule definition phase to balance the design request and litho Optical proximity correction(OPC) complexity. Among the layout pattern clips (~100K layout patterns in the case study), a few will fail after OPC at litho simulation. Design rule definition team need to debug the violations to develop either a new OPC recipe or a new set of rule candidates to remove the specific violations. We have demonstrated a full Machine learning capability and applied in an advanced node process design rule definition. The paper will cover details:
1) extremely unbalanced dataset from lithography simulation
2) Machine learning model selection and training
3) Design rule extraction from Machine learning model
4) An random forest based rule generation and rule ranking to produce rule candidates
The paper further elaborate future direction to drive to symbolic learning to give a reasoning layer from general machine learning model to symbolic rule definition.
1) extremely unbalanced dataset from lithography simulation
2) Machine learning model selection and training
3) Design rule extraction from Machine learning model
4) An random forest based rule generation and rule ranking to produce rule candidates
The paper further elaborate future direction to drive to symbolic learning to give a reasoning layer from general machine learning model to symbolic rule definition.