SURGE - Surrogate Unified Robust Generation Engine

Le résumé fourni par la source

SURGE is a surrogate modeling framework for scientific workflows that integrates Scientific Machine Learning (SciML) and AutoML features, uncertainty quantification (UQ), and MLOps-grade provenance in a single declarative pipeline. It unifies data generation and ingestion, an extensible registry of model adapters (classical, neural, probabilistic, and ensemble), held-out and cross-validated evaluation with UQ, automated hyperparameter optimization, structured artifact and lineage tracking, diagnostic visualization, and portable inference and deployment. Configuration-as-code specifications parameterize a composable, end-to-end surrogate development cycle and emit machine-readable provenance. This standardizes methodology, reducing domain-expert adaptation effort and accelerating the surrogate development cycle, while exposing APIs consumable by AI-agentic orchestration. By combining an automated, composable surrogate pipeline with an extensible adapter registry, disciplined artifact lineage, and agent-ready interfaces, SURGE provides a robust substrate for training, validating, comparing, and operationalizing surrogate models across scientific domains.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.