Explainable machine learning of the underlying physics of high-energy particle collisions

ML Generative Adversarial Network learns internal workings of QCD - the Altarelli-Parisi splitting function - using information encoded in final state particle distributions. We present an implementation of an explainable and physics-aware machine learning model capable of inferring the underlying physics of high-energy particle collisions using the information encoded in the energy-momentum four-vectors of the final state particles. We demonstrate the proof-of-concept of our White Box AI approach using a Generative Adversarial Network (GAN) which learns from a DGLAP-based parton shower Monte Carlo event generator. We show, for the first time, that our approach leads to a network that is able to learn not only the final distribution of particles, but also the underlying parton branching mechanism, i.e. the Altarelli-Parisi splitting function, the ordering variable of the shower, and the scaling behavior. 

More reading: https://arxiv.org/abs/2012.06582 

RNC Program. Work supported by LBNL's LDRD program.

CONTExt aware ML analysis of nuclear radiation sources

Advanced methods of data fusion from mobile detectors equipped with multiple sources of data (including LIDAR and video feeds) can advise on nuclear safety and security . Radiation mapping can be used to improve safety at sites with radioactive sources (such as power plants or hospitals), enforce non-proliferation agreements, or guide environmental cleanup and disaster response. Scientists at Berkeley Lab have created multi-sensor systems that can map nuclear radiation in 3D in real time and are now testing how to integrate their system with robots that can autonomously investigate radiation areas. New ways of monitoring radiation within city limits via a network of distributed sensors are being developed.

More reading: https://anp.lbl.gov 

Applied Nuclear Science

The information content of jet quenching and machine learning assisted observable design

ML guides future directions of the experimental program focused on the understanding microscopic structure of the quark-gluon plasma created in heavy-ion collisions at collider experiments (RHIC and LHC).  Jets produced in high-energy heavy-ion collisions are modified compared to those in proton-proton collisions due to their interaction with the deconfined, strongly-coupled quark-gluon plasma (QGP). In this work, we employ machine learning techniques to identify important features that distinguish jets produced in heavy-ion collisions from jets produced in proton-proton collisions. We formulate the problem using binary classification and focus on leveraging machine learning in ways that inform theoretical calculations of jet modification.

More reading: https://arxiv.org/abs/2111.14589 

RNC Program. Work supported by LBNL's LDRD program.

ML used to improve performance of ion sources and tracking detectors

NSD scientists are developing ML aided optimizations for instrumentation upstream and downstream of particle accelerators: the cases of VENUS and GRETA.

The study of atomic nuclei is central to our understanding of the world around us.  Comprising 99.9% of the visible matter in the universe nuclei are, in multiple aspects, central to fundamental questions in physics, such as our understanding of the origin of the elements and how complex many-body quantum systems organize. Their properties depend sensitively on the number of protons (Z) and neutrons (N). Understanding nuclear properties, their role within the cosmos, and more broadly their application for society requires measurements on elements and isotopes far from β stability.

More reading: https://cyclotron.lbl.gov/ionsources/venus  https://greta.lbl.gov 

Nuclear Spectroscopy, ML project funded by DOE-NP

Machine learning-based jet and event classification at the Electron-Ion Collider with applications to hadron structure and spin physics

ML guides future directions of the experimental program accessible with the new Electron-Ion Collider at Brookhaven National Laboratory.  We explore machine learning-based jet and event identification at the future Electron-Ion Collider (EIC). We study the effectiveness of machine learning-based classifiers at relatively low EIC energies, focusing on (i) identifying the flavor of the jet and (ii) identifying the underlying hard process of the event. We propose applications of our machine learning-based jet identification in the key research areas at the future EIC and current Relativistic Heavy Ion Collider program, including enhancing constraints on (transverse momentum dependent) parton distribution functions, improving experimental access to transverse spin asymmetries, studying photon structure, and quantifying the modification of hadrons and jets in the cold nuclear matter environment in electron-nucleus collisions.

More reading: https://arxiv.org/abs/2210.06450 

RNC & NSD-Theory programs within NSD.

IMproved insights from Bayesian uncertainty quantification

Bayesian inference and its uncertainty quantification when comparing a theoretical or computational model to observations gains strong attention in multiple domains. NSD scientists working have proposed building a common framework for multi-discipline effort to formulate a framework for quantifying the uncertainty for physics research (including applied physics problems) that uses Bayesian inference. The computationally intensive inverse problems in NP include areas of neutrino physics, properties of quark gluon plasma, radiological mapping of the environment, gamme particle tracking algorithms but also a number of areas in highe energy physics. With the increased use of data-driven modeling proper understanding of uncertainties in such analyses is critical to conducting a sound scientific investigation.

RNC, ANP, Neutrinos, Accelerator Based Low Energy Research programs

ML project funded by DOE-NP

Is infrared-collinear safe information all you need for jet classification?

Next level of exploration of effects of hot and dense quark-gluon plasma on jets propagating within its volume - new insight for experimental program at RHIC and the LHC. Machine learning-based jet classifiers are able to achieve impressive tagging performance in a variety of applications in high energy and nuclear physics. However, it remains unclear in many cases which aspects of jets give rise to this discriminating power, and whether jet observables that are tractable in perturbative QCD such as those obeying infrared-collinear (IRC) safety serve as sufficient inputs. In this article, we introduce a new classifier, Jet Flow Networks (JFNs), in an effort to address the question of whether IRC unsafe information provides additional discriminating power in jet classification.

More reading: https://arxiv.org/abs/2305.08979 

RNC Program. Work supported by LBNL's LDRD program.