Welcome to BFBrain’s documentation!
BFBrain is a Python library for training Bayesian neural networks to approximate bounded-from-below conditions for multiscalar potentials in quantum field theory. In many multiscalar theories, determining if a given point in parameter space is bounded-from-below is both prohibitively computationally expensive to do numerically and intractable to resolve into symbolic conditions. BFBrain works to resolve this issue by encoding approximate bounded-from-below conditions in a Bayesian neural network, which can be used as portable, efficiently-computable boundedness-from-below conditions in parameter space scans for BSM physics studies. In [1], our paper introducing BFBrain, we have found that this methodology can significantly outperform more conventional methods of approximating bounded-from-below conditions while being theoretically applicable to any renormalizable scalar potential and exhibiting robust uncertainty determination which can inform a user when a model’s predictions should be trusted. The methods of the BFBrain package are designed to allow rapid implementation of analyses with only a few lines of code, but remain open to a high degree of user customization.
Check out the Usage section for further information, including how to install the project and a simple example of how to get started with Quickstart. For a more detailed introduction to the software with a step-by-step tutorial, see Tutorial and User Guide. For a full API reference, see BFBrain