Regression Algorithms Evaluation for Analysis of Crosstalk in High-Speed Digital System

초록

As technology advances, processor speeds are increasing at a rapid pace and digital systems require a significant amount of data bandwidth. As a result, careful consideration of signal integrity is required to ensure reliable and high-speed data processing. Crosstalk has become a vital area of research in signal integrity for electronic packages, mainly because of the high level of integration. Analytic formulas were analyzed in this study to identify the features that can predict crosstalk in multi -conductor transmission lines. Through the analysis, five variables were found and obtained a dataset consisting of 302,500, data points. The study evaluated the performance of various regression models for optimization via automatic machine learning by comparing the machine learning predictions with the analytic solution. Extra tree regression consistently outperformed other algorithms, with coefficients of determination exceeding 0.9 and root mean square logarithmic errors below 0.35. The study also notes that different algorithms produced varied predictions for the two metrics.

키워드