Open Source Machine Learning Model Safety Assurance for Embedded Edge Systems

EasyChair Preprint 12739

Abstract

Machine learning (ML) model advancement is moving forward at a rapid pace compared to the safety and security critical system development life cycle.  The behavior of the ML model depends on the data and software implementation of the algorithm.  To leverage these advancements while keeping the system compliant for safety critical applications, this paper reviews the challenges of using open source ML models for safety critical applications and proposes metrics and workflow to improve model assurance for deployment in edge systems.\

Keyphrases: Explainability, Safety Assurance, Security, data bias, data variance, embedded edge, machine learning, model validation, open source, safety

@booklet{EasyChair:12739,
  author    = {Ching Hu and S.R. Prakash},
  title     = {Open Source Machine Learning Model Safety Assurance for Embedded Edge Systems},
  howpublished = {EasyChair Preprint 12739},
  year      = {EasyChair, 2024}}