[Paper Review] HepData and JetWeb: HEP data archiving and model validation

A. G. Buckley, J. M. Butterworth+5 more|ArXiv.org|May 4, 2006

Advanced Data Storage Technologies1 references5 citations

TL;DR

This paper presents CEDAR, a unified system integrating HepData and JetWeb to streamline high-energy physics (HEP) model tuning and validation. By using a standardized HepML XML format for experimental and generator data, the system enables automated, reproducible comparisons between Monte Carlo simulations and experimental observables, significantly improving the robustness and transparency of event generator parameter tuning for LHC physics analyses.

ABSTRACT

The CEDAR collaboration is extending and combining the JetWeb and HepData systems to provide a single service for tuning and validating models of high-energy physics processes. The centrepiece of this activity is the fitting by JetWeb of observables computed from Monte Carlo event generator events against their experimentally determined distributions, as stored in HepData. Caching the results of the JetWeb simulation and comparison stages provides a single cumulative database of event generator tunings, fitted against a wide range of experimental quantities. An important feature of this integration is a family of XML data formats, called HepML.

Motivation & Objective

• To address the challenge of inconsistent and non-reproducible tuning of Monte Carlo event generators in high-energy physics.
• To create a centralized, standardized system for storing and comparing experimental data with simulated observables.
• To improve model validation by enabling simultaneous fitting of multiple experimental observables across diverse processes.
• To establish a common data format (HepML) for experimental results and generator configurations to ensure interoperability.
• To support the development of a sustainable, community-curated infrastructure for HEP software and data.

Proposed method

• Integrating HepData’s experimental reaction data repository with JetWeb’s simulation and fitting engine to create a unified tuning service.
• Defining a family of XML-based data formats (HepML) for representing experimental data and event generator configurations using XML Schema.
• Using XSLT and XPath to dynamically generate web-based visualizations and output formats from HepML records.
• Migrating HepData’s legacy database to a relational MySQL schema with a new Java object model for improved data management.
• Enhancing JetWeb to directly query HepData, eliminating data conversion errors and enabling real-time updates.
• Developing HZTool (Fortran) and its C++ replacement Rivet for computing observable distributions from Monte Carlo events, with generator-agnostic interfaces.

Experimental results

Research questions

• RQ1 How can experimental HEP data and Monte Carlo simulation results be systematically and reproducibly compared across diverse processes?
• RQ2 What standardized data format is needed to ensure interoperability between experimental data archives and event generator configurations?
• RQ3 Can a unified system reduce the risk of non-physical predictions arising from tuning to limited observables?
• RQ4 How can the integration of HepData and JetWeb improve the transparency and reusability of event generator tuning?
• RQ5 What infrastructure is required to support community-driven, long-term validation of HEP simulation models?

Key findings

• The HepML family of XML formats has been defined and released for public comment, with initial schemas for HepData records and event generator configurations.
• HepData’s legacy BDMS database has been successfully migrated to a relational MySQL database with a new Java object model, improving data integrity and access.
• A proof-of-concept HepData website using XSLT transformations of HepML records is now available, demonstrating dynamic data presentation.
• JetWeb has been updated to use HepML for generator parameter storage, support AIDA-compliant plotting, and simplify the addition of new event generator models.
• The HZTool library is being succeeded by Rivet, a C++-based, object-oriented replacement designed for better maintainability and integration with modern generators.
• CEDAR is on track to deliver globally validated, reproducible event generator tunings for LHC physics analyses through its integrated, community-curated infrastructure.