Efficient Bayesian Yield Analysis and Optimization with Active Learning
TimeThursday, July 14th4:50pm - 5:10pm PDT
Location3000, Level 3
Event Type
Research Manuscript
Design Verification and Validation
DescriptionIn this work, a uniform framework to solve yield estimation and optimization simultaneously is proposed. A novel efficient Bayesian yield analysis framework, BYA, is firstly proposed by deriving a Bayesian estimation of the yield and introducing an activate learning based on the integration of entropy reduction. A convolutional optimization is also proposed to solve the entropy reduction problem efficiently. We then further extend BYA for yield optimization. Our method demonstrates up to 5x speed-up in yield estimation and up to 10x speed-up in yield optimization respectively compared with the SOTA methods.