ABOT toolbox V2
The toolbox enables cost-efficient optimization of manufacturing process design parameters. It achieves this through data-efficient methodologies for sequential experiment design, streamlining and enhancing optimization processes..
Description:
This toolbox has significantly advanced Bayesian optimization methods, creating a versatile toolbox for various sectors. It optimizes high-cost production, logistics, and healthcare simulations. The project developed an advanced simulation model for different decision variables and implemented state-of-the-art methods for mixed input spaces. Positive benchmarking results show it's ready for industry use, offering an efficient alternative to traditional trial-and-error methods, especially in high-cost scenarios.
In this latest version, the ABOT toolbox v2 has been upgraded to use state-of-the-art Python libraries like BOTorch, which relies on PyTorch for machine learning. The latest Matlab simulator from Flanders Make now interacts smoothly with this Python framework. Additionally, the adhesive bonding optimization problem and simulator have been fully integrated into BOTorch, allowing it to handle continuous, discrete, and categorical parameters. The toolbox also provides an interactive way to run experiments for various real-life scenarios provided by the simulator.
Please view the ‘Getting started guide’ here
More Information:
Enrichment project: Data-efficient Machine learning for Smart Design of Experiments
Responsible Research Lead: Ivo Couckuyt (PI), Inneke Van Nieuwenhuyse (PI) and Jeroen Jordens (Senior research engineer)
Authors: Sebastian Rojas Gonzalez
Programming language: Python
Categories: Bayesian Optimization, Simulation Integration, Mixed Input Spaces