A Software Framework for Automated Behavioral Modeling of Electronic Devices [Application Notes]

Abstract

Abstract-As the fabrication of prototypes is very costly, system-level computer simulations have become commonplace alternatives for the design of electronic devices and systems. However, due to the computational cost of these simulations, the use of behavioral modeling techniques has become indispensable. Behavioral models can act as a surrogate for the expensive simulations, and effectively speed-up the simulations without sacrificing accuracy. Such models are compact and cheap to evaluate, and have proven very useful for tasks such as optimization, design space exploration, prototyping, and sensitivity analysis. Consequently, there is great interest in techniques that facilitate the construction of behavioral models, while minimizing the computational cost and maximizing model accuracy. This paper explores how such models can be calculated in an automated way by means of a unified software framework that integrates adaptive modeling and adaptive sampling methods. It generates global behavioral models that are accurate and valid over the design space of interest while minimizing the number of electromagnetic simulations. By placing a strong focus on adaptivity, flexibility, and self-tuning the burden on the designer is relieved and the total modeling turn-around time is reduced.

Adaptive classification under computational budget constraints using sequential data gathering

“…These methods are typically complemented with an input-based method to ensure not too much focus is put on the exploitation of the acquired knowledge. All listed algorithms include an input-based component to ensure exploration as well, and are popular methods for building designs with several successful applications [24,29,40,30,41].…”

Adaptive modeling and sampling methodologies for Internet of Things applications

“…The toolbox has been used previously in several successful engineering applications in several domains including electronics [7].…”

The SUMO toolbox: A tool for automatic regression modeling and active learning

“…As a simple example of a problem that was successfully modeled using the SUMO toolbox, we present the following application from electro-magnetism (code courtesy of Robert Lehmensiek [13]): a 3-dimensional simulation model that computes the scattering parameters for a step discontinuity in a rectangular waveguide [14]. The inputs consists of the input frequency, the gap height and the gap length.…”