High-Speed Link Design Optimization Using Machine Learning SVR-AS Method

This paper proposes a novel and fast constrained design optimization method based on support vector regression-active subspace method. The proposed optimization method calculates a linear combination of original design parameters named active variable as a low-dimensional representation of high-dimensional design space to transform the non-linear constraint into a reduced linear constraint for optimization problems, which successfully derives a simplified and mathematically solvable equation. A complex high-speed link with 16-dimensional design parameters is utilized to verify this method and results show that the proposed method can efficiently find the optimal design structures compared to interior-point method.