Robust Analysis for Resilient AI System

2025-09-07

AI Summary

In manufacturing industrial internet (MII) systems, operation-risk-induced data anomalies severely undermine the reliability of conventional statistical methods. To address this, we propose DPD-Lasso—a novel regression framework integrating density power divergence (DPD)-based robust estimation with Lasso variable selection, specifically designed for high-noise, anomaly-contaminated industrial data. We develop an efficient iterative algorithm with guaranteed convergence to overcome computational bottlenecks. Empirical validation on an aerosol jet printing experimental platform demonstrates that DPD-Lasso maintains stable prediction performance and accurate risk quantification under both clean and anomaly-contaminated conditions, significantly enhancing the reliability of resilience assessment for industrial AI systems. The method establishes a new paradigm for robust statistical modeling in MII contexts—offering both interpretability and scalability while preserving theoretical rigor and practical efficacy.

Technology Category

• Intelligent Robots: State Estimation
• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Operational hazards in Manufacturing Industrial Internet (MII) systems generate severe data outliers that cripple traditional statistical analysis. This paper proposes a novel robust regression method, DPD-Lasso, which integrates Density Power Divergence with Lasso regularization to analyze contaminated data from AI resilience experiments. We develop an efficient iterative algorithm to overcome previous computational bottlenecks. Applied to an MII testbed for Aerosol Jet Printing, DPD-Lasso provides reliable, stable performance on both clean and outlier-contaminated data, accurately quantifying hazard impacts. This work establishes robust regression as an essential tool for developing and validating resilient industrial AI systems.

Problem

• Robust regression for contaminated industrial data analysis
• Overcoming computational bottlenecks in resilient AI systems
• Quantifying hazard impacts on manufacturing AI performance

Innovation

• DPD-Lasso integrates Density Power Divergence with Lasso
• Efficient iterative algorithm overcomes computational bottlenecks
• Provides reliable performance on clean and outlier data