Exploring the Impact of Parametric Variability on Eye Diagram of On-Chip Multi-walled Carbon Nanotube Interconnects using Fast Machine Learning Techniques

In this paper, an artificial neural network (ANN) based metamodel is developed for estimating the eye diagram characteristics of on-chip multi-walled carbon nanotube (MWCNT) interconnect networks subject to parametric variability. The proposed ANN metamodel is trained using a prior knowledge input with source difference (PKID) formulation. In this PKID formulation, the eye diagram characteristics predicted by the compact but approximate equivalent single conductor (ESC) representation of the MWCNT conductors is used as prior knowledge to accelerate the learning of the ANN. Consequently, the proposed ANN metamodel can be trained at far smaller computational time costs than conventional ANN metamodels and existing multi-fidelity approaches without any discernable loss in accuracy.