Check out the new tools and examples in BAND Framework v0.4!
See the BAND Manifesto in Journal of Physics G! (or on arXiv)
Nuclear physicists seek an accurate description of the properties of atomic nuclei, collisions between nuclei, and extreme environments such as the first few seconds of our universe or the interior of a neutron star. These situations involve many particles interacting through complex forces. They’re each described by a number of different models: these typically do a good job of explaining the results of experiments that have already occurred. The models don’t do as well predicting what will happen in future experiments or in environments that are inaccessible here on Earth.
The Bayesian Analysis of Nuclear Dynamics (BAND) Framework will use advanced statistical methods to produce forecasts for as-yet-unexplored situations that combine nuclear-physics models in an optimal way. These will be more reliable than the predictions of any individual model. BAND’s forefront computer codes will be widely available and will facilitate the design of nuclear-physics experiments that can deliver the largest gain in understanding. The adoption of BAND’s tools in other sciences dealing with “model uncertainty” could spur broad scientific innovation. Undergraduate and graduate students working on BAND will gain a broad range of technical skills in data science, machine learning, nuclear physics, and high-performance computing.
BAND Code of Conduct
BAND is committed to fostering a safe, diverse, inclusive, and equitable environment that values mutual respect and personal integrity. Therefore, all participants in BAND activities will conduct themselves in a welcoming and professional manner, treating one another respectfully and considerately. This is particularly important in interdisciplinary work that brings together collaborators from different scientific backgrounds and cultures. Creating such a culture–one that is collegial, inclusive, and professional–is the responsibility of all collaboration members and meeting participants.
Those participating in BAND activities shall not discriminate against, harass, or bully others. If you observe inappropriate comments or actions and personal intervention seems appropriate and safe, you are encouraged to intervene in ways that are, insofar as it is possible, considerate of all parties involved. BAND members and meeting participants will not discriminate against, act in an inappropriate way, or make inappropriate statements related to aspects of each other’s identity such as age, race, ethnicity, perceived or actual social class, sexual orientation, gender identity, gender expression, marital status, nationality, political affiliation, religion, ability status, and educational background. No kind of harassment (including sexual harassment) or bullying will be tolerated.
Bullying is unwelcome, aggressive behavior involving the use of influence, threat, intimidation, or coercion to dominate others. Harassment includes but is not limited to: inappropriate or intimidating behavior and language, unwelcome jokes or comments, unwanted touching or attention, offensive images, photography without permission, and stalking. Sexual harassment is unwelcome sexual advances, requests for sexual favors, and other verbal or physical conduct of a sexual nature that creates an intimidating, hostile, or offensive environment.
Violations of this code of conduct policy should be reported to meeting organizers and/or the BAND “Community Leaders”: Daniel Phillips (phillid1@ohio.edu), Dick Furnstahl (furnstahl.1@osu.edu), Özge Sürer (surero@miamioh.edu), Frederi Viens (fv15@rice.edu), and Stefan Wild (wild@lbl.gov). Sanctions may range from verbal warning, to ejection from the meeting, to a suspension from future BAND activities, to the notification of appropriate local authorities. Retaliation for complaints of inappropriate conduct will not be tolerated and will result in an escalated sanction.
On December 17 and 18 members of BAND met with several scientists interested in the curation and evaluation of Nuclear Data to discuss topics of mutual interest. A list of presentations at that meeting, togehter with .pdf of some of the slides presented, can be found here.
December 14, 2024BAND co-PI Stefan Wild presented work, An Active Learning Performance Model for Parallel Bayesian Calibration of Expensive Simulations with Ozge Surer at the NeurIPS Workshop on Bayesian Decision-Making and Uncertainty in Vancouver.
October 21, 2024The BAND framework was well represented at the recently concluded Fall Meeting of the Division of Nuclear Physics of the American Physical Society. BAND Senior Investigator Filomena Nunes presented as part of a career development panel for early-career researchers. BAND Researcher Pablo Giuliani ran the "Lifting the Shadows" session, which aimed to promote more inclusive nuclear-physics workspaces. And BAND researcher Kyle Godbey chaired a session in which BAND Fellow Christal Martin and researcher Oleh Savchuk both presented. BAND members Daniel Phillips and Manuel Catacorarios also gave talks at the meeting.
October 16, 2024BAND Senior Investigator Dick Furnstahl is the recipient of the 2025 Herman Feshbach Prize in Theoretical Nuclear Physics. Furnstahl was cited "For found foundational contributions to calculations of nuclei, including...using Bayesian methods to quantify the uncertainties in effective field theory predictions of nuclear observables."
October 2, 2024v0.4 of the BAND Framework is out! This version includes updates of existing BAND packages such as surmise, ROSE, and Taweret. It also includes new capabilities, such as PUQ, a Python package that employs novel experimental design techniques with intelligent selection criteria, and nsat, code that implements a Bayesian mixture model approach to quantifying the empirical nuclear saturation point.
September 9, 2024BAND members Pablo Giuliani, Kyle Godbey, and Witek Nazarewicz collaborted with statistician Vojtech Kejzlar to show how "Principal Component Analysis" can be used to ensure that only independent---or even better, orthogonal---models get included in forecasts that combine models using Bayesian methods. Their paper is out today in Physical Review Research.