First Light Fusion – Data Science Toolkit
The First Light Fusion Data Science Toolkit is an integrated software suite designed to accelerate simulation studies in physics and engineering using the latest Probabilistic Machine Learning (ML) and Artificial Intelligence (AI) techniques.
Implemented as a pure Python library, the suite offers a full range of functionalities with a straightforward and well-documented API. It ensures reliable and rapid improvements in results by intelligently driving simulation codes through hundreds of runs. Latest parameter exploration, optimisation and inference tools enable the user to rapidly evaluate a full range of options and to fully understand their model.
It is a proven and versatile technology, having been used extensively in the design of First Light’s Amplifiers, Inertial Fusion Targets and Pulsed Power Drivers, making it an ideal solution to accelerate simulation-driven design work in high tech industries.
Features
- A robust yet simple interface to simulation codes, allowing it to drive any command-line executable or Python scriptable simulation model.
- A job-marshaller executes very large ensembles of simulations (1000+ runs) distributed on high performance computing platforms (on premises or cloud).
- Rapid model exploration and scoping with flexible, state-of-the-art stratified space filling designs and active learning.
- Extensive surrogating support, including (Deep Kernel) Gaussian Process and Neural Network based surrogates, to enable rapid response surface modelling.
- Local and global sensitivity analysis tools to identify the most impactful input parameters.
- Robust prediction and reliability analysis tools to build confidence in simulated predictions.
- Powerful optimisation tools based on Bayesian Optimisation and Evolution Strategies allow both local and global optima to be explored quickly.
Contact
Please contact software@firstlightfusion.com to discuss your use case & how the First Light Fusion Data Science Toolkit can help.