Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:M3-2
Session:
Number:M3-2-3
Random Forest Model for Silicon-to-SPICE Gap and FinFET Design Attributes Identification
Hyosig Won, Katsuhiro Shimazu ,
pp.299-302
Publication Date:2016/7/10
Online ISSN:2188-5079
DOI:10.34385/proc.61.M3-2-3
PDF download (1.9MB)
Summary:
Random forest model was applied to analyze design attributes influence on Silicon-to-SPICE(S2S) gap. In order to have enough model accuracy to discuss S2S gap, scaled learning data was used with random design attributes count to build each tree in forest model. From the improved model, indices so-called `importance' and newly defined `impact' can be extracted to identify significant design attributes which determine S2S gap. The identified design attributes classify S2S gap well and show clear trend of it. Finally the key FinFET structures can be identified as the representative layout structure to cause large S2S gap.