-
Training Data Selection for Accuracy and Transferability of Interatomic Potentials
Authors:
David Montes de Oca Zapiain,
Mitchell A. Wood,
Nicholas Lubbers,
Carlos Z. Pereyra,
Aidan P. Thompson,
Danny Perez
Abstract:
Advances in machine learning (ML) techniques have enabled the development of interatomic potentials that promise both the accuracy of first principles methods and the low-cost, linear scaling, and parallel efficiency of empirical potentials. Despite rapid progress in the last few years, ML-based potentials often struggle to achieve transferability, that is, to provide consistent accuracy across co…
▽ More
Advances in machine learning (ML) techniques have enabled the development of interatomic potentials that promise both the accuracy of first principles methods and the low-cost, linear scaling, and parallel efficiency of empirical potentials. Despite rapid progress in the last few years, ML-based potentials often struggle to achieve transferability, that is, to provide consistent accuracy across configurations that significantly differ from those used to train the model. In order to truly realize the promise of ML-based interatomic potentials, it is therefore imperative to develop systematic and scalable approaches for the generation of diverse training sets that ensure broad coverage of the space of atomic environments. This work explores a diverse-by-construction approach that leverages the optimization of the entropy of atomic descriptors to create a very large ($>2\cdot10^{5}$ configurations, $>7\cdot10^{6}$ atomic environments) training set for tungsten in an automated manner, i.e., without any human intervention. This dataset is used to train polynomial as well as multiple neural network potentials with different architectures. For comparison, a corresponding family of potentials were also trained on an expert-curated dataset for tungsten. The models trained to entropy-optimized data exhibited vastly superior transferability compared to the expert-curated models. Furthermore, while the models trained with heavy user input (i.e., domain expertise) yield the lowest errors when tested on similar configurations, out-sample predictions are dramatically more robust when the models are trained on a deliberately diverse set of training data. Herein we demonstrate the development of both accurate and transferable ML potentials using automated and data-driven approaches for generating large and diverse training sets.
△ Less
Submitted 5 May, 2022; v1 submitted 24 January, 2022;
originally announced January 2022.
-
The ProtoDUNE-SP LArTPC Electronics Production, Commissioning, and Performance
Authors:
D. Adams,
M. Bass,
M. Bishai,
C. Bromberg,
J. Calcutt,
H. Chen,
J. Fried,
I. Furic,
S. Gao,
D. Gastler,
J. Hugon,
J. Joshi,
B. Kirby,
F. Liu,
K. Mahn,
M. Mooney,
C. Morris,
C. Pereyra,
X. Pons,
V. Radeka,
E. Raguzin,
D. Shooltz,
M. Spanu,
A. Timilsina,
S. Tufanli
, et al. (9 additional authors not shown)
Abstract:
The ProtoDUNE-SP detector is a large-scale prototype of the Single-Phase (SP) Liquid Argon Time Projection Chamber (LArTPC) design proposed for the Deep Underground Neutrino Experiment (DUNE). 15,360 LArTPC wires are instrumented with low electronic noise pre-amplifier and digitization ASICs integrated into Front End Motherboards (FEMBs) operating at cryogenic temperature within the cryostat. The…
▽ More
The ProtoDUNE-SP detector is a large-scale prototype of the Single-Phase (SP) Liquid Argon Time Projection Chamber (LArTPC) design proposed for the Deep Underground Neutrino Experiment (DUNE). 15,360 LArTPC wires are instrumented with low electronic noise pre-amplifier and digitization ASICs integrated into Front End Motherboards (FEMBs) operating at cryogenic temperature within the cryostat. The large number of electronics channels and high performance specifications required a large-scale production electronics quality control effort, careful installation into Anode Plane Assemblies (APAs), and rigorous detector commissioning. This successful collaboration-wide effort achieved a working LArTPC electronics channel percentage of 99.7% (15,318 of 15,360 channels in total), whose operating performance exceeded expectations. We summarize the ProtoDUNE-SP cold electronics design and quality control, installation, and commissioning efforts that enabled this excellent electronics performance.
△ Less
Submitted 30 April, 2020; v1 submitted 5 February, 2020;
originally announced February 2020.
-
Dirac equation in (1+3) and (2+2) dimensions
Authors:
J. A. Nieto,
C. Pereyra
Abstract:
We develop a systematic method to derive the Majorana representation of the Dirac equation in (1+3)-dimensions. We compare with similar approach in (2+2)-dimensions . We argue that our formalism can be useful to have a better understanding of possible Majorana fermions.
We develop a systematic method to derive the Majorana representation of the Dirac equation in (1+3)-dimensions. We compare with similar approach in (2+2)-dimensions . We argue that our formalism can be useful to have a better understanding of possible Majorana fermions.
△ Less
Submitted 18 July, 2013; v1 submitted 23 May, 2013;
originally announced May 2013.
