jetto_mobo documentation#
Multi-objective Bayesian optimisation for JETTO in Python
jetto_mobo brings together tools for optimisation (design), running JETTO (simulation), and for evaluating the optimisation results (evaluation)#
jetto_mobo is a Python package for performing multi-objective Bayesian optimisation (BO) in the JETTO plasma modelling code.
It is a slim wrapper around BoTorch, a library for Bayesian optimisation in PyTorch.
JETTO is run using Singularity/Apptainer.
The project’s aim is to provide a plug-and-play package for performing plasma scenario optimisation in JETTO. This means less faffing around with PyTorch tensors and boilerplate, allowing users to leverage the power of model-based optimisation without requiring familiarity with machine-learning libraries.
We want to provide a simple drop-in replacement that allows users to easily switch to using a more efficient and interpretable optimisation algorithm.