Dear UQWorld Community,
Since its release, UQLab has grown into a reference platform for uncertainty quantification, recently reaching nearly 10,000 users in 100 countries. Today, we are announcing the release of UQLab version 2.2.
This update introduces a new module developed specifically for the rigorous, repeatable, and objective benchmarking of surrogate models on engineering problems and datasets. This release marks an important step toward standardizing the evaluation of surrogate models, providing a framework to foster transparent and reproducible comparisons across the UQ field.
Key features of the new benchmarking module include:
-
Statistical robustness through automated replications.
-
Flexible performance metrics, including user-defined measures.
-
Interactive reporting for efficient data filtering and interpretation.
-
Advanced visualization and aggregation tools for ranking methods.
-
High-performance computing (HPC) support for large-scale benchmarking studies.
-
An extensive library of precomputed results, allowing for immediate comparison with state-of-the-art methods.
-
Seamless import/export capabilities for integration with external datasets and tools.
Additional Improvements: Beyond the benchmarking module, version 2.2 also brings:
-
Enhanced capabilities in the Active Learning Reliability module.
-
Updates to the HPC Dispatcher module for improved performance and usability.
For full technical details, please consult the Release Notes. You can download UQLab 2.2 at www.uqlab.com.
We welcome your technical questions and discussions regarding these new tools in this thread.
The UQLab Development Team