Extensive studies of the neutron star equation of state from the deep learning inference with the observational data augmentation

Fujimoto, Yuki; Fukushima, Kenji; Murase, Koichi

doi:10.1007/jhep03(2021)273

Cited by 46 publications

(38 citation statements)

References 100 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…At asymptotically high densities of ρ 50 ρ 0 , perturbative QCD calculations converge and provide reliable results [51][52][53][54][55][56][57]. At intermediate densities of ρ ∼ [2 − 10]ρ 0 , however, there are still no reliable QCD predictions [58]. To derive the EOS from QCD in the intermediatedensity region, one needs to develop non-perturbative approaches, such as the Monte Carlo simulation of QCD on a lattice (lattice QCD), but the application of these methods to systems at finite densities is hindered by the notorious sign problem; see, e.g., Ref.…”

Section: Introductionmentioning

confidence: 99%

Translating Neutron Star Observations to Nuclear Symmetry Energy via Deep Neural Networks

Krastev

2022

Galaxies

View full text Add to dashboard Cite

One of the most significant challenges involved in efforts to understand the equation of state of dense neutron-rich matter is the uncertain density dependence of the nuclear symmetry energy. In particular, the nuclear symmetry energy is still rather poorly constrained, especially at high densities. On the other hand, detailed knowledge of the equation of state is critical for our understanding of many important phenomena in the nuclear terrestrial laboratories and the cosmos. Because of its broad impact, pinning down the density dependence of the nuclear symmetry energy has been a long-standing goal of both nuclear physics and astrophysics. Recent observations of neutron stars, in both electromagnetic and gravitational-wave spectra, have already constrained significantly the nuclear symmetry energy at high densities. The next generation of telescopes and gravitational-wave observatories will provide an unprecedented wealth of detailed observations of neutron stars, which will improve further our knowledge of the density dependence of nuclear symmetry energy, and the underlying equation of state of dense neutron-rich matter. Training deep neural networks to learn a computationally efficient representation of the mapping between astrophysical observables of neutron stars, such as masses, radii, and tidal deformabilities, and the nuclear symmetry energy allows its density dependence to be determined reliably and accurately. In this work, we use a deep learning approach to determine the nuclear symmetry energy as a function of density directly from observational neutron star data. We show, for the first time, that artificial neural networks can precisely reconstruct the nuclear symmetry energy from a set of available neutron star observables, such as masses and radii as measured by, e.g., the NICER mission, or masses and tidal deformabilities as measured by the LIGO/VIRGO/KAGRA gravitational-wave detectors. These results demonstrate the potential of artificial neural networks to reconstruct the symmetry energy and the equation of state directly from neutron star observational data, and emphasize the importance of the deep learning approach in the era of multi-messenger astrophysics.

show abstract

Section: Introductionmentioning

confidence: 99%

Translating Neutron Star Observations to Nuclear Symmetry Energy via Deep Neural Networks

Krastev

2022

Galaxies

View full text Add to dashboard Cite

show abstract

“…At asymptotically high densities of ρ > ∼ 50ρ 0 perturbative QCD calculations converge and provide reliable results [51][52][53][54][55][56][57]. At intermediate densities of ρ ∼ [2 − 10]ρ 0 , however, there are still no reliable QCD predictions [58]. To derive the EOS from QCD in the intermediate density region, one needs to develop non-perturbative approaches, such as the Monte-Carlo simulation of QCD on a lattice (lattice QCD), but the application of these methods to systems at finite densities is hindered by the notorious sign problem, see, e.g.…”

Section: Introductionmentioning

confidence: 99%

“…[85]. Despite the significant effort to extract the genuine EOS from the NS astrophysical data, it is still unclear what the true dense matter EOS should look like mainly due to the uncertainties of the assumed prior distributions in the Bayesian analyses [58]. Therefore, the need arises of alternative approaches to construct the model independent EOS.…”

Section: Introductionmentioning

confidence: 99%

“…In previous works [125,126], we also pioneered the use of DL methods, specifically Convolutional Neural Network (CNN) [127] algorithms, for detection and inference of GW signals from binary neutron star (BNS) mergers embedded in both Gaussian and realistic LIGO noise. Several studies have also explored the DL approach as a tool to extract the dense matter EOS from NS observations [58,[128][129][130][131].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Translating neutron star observations to nuclear symmetry energy via artificial neural networks

Krastev

2021

Preprint

View full text Add to dashboard Cite

One of the most significant challenges involved in efforts to understand the equation of state of dense neutron-rich matter is the uncertain density dependence of the nuclear symmetry energy. In particular, the nuclear symmetry energy is still rather poorly constrained, especially at high densities. On the other hand, detailed knowledge of the equation of state is critical for understanding of many important phenomena in the nuclear terrestrial laboratories and the cosmos. Because of its broad impact, pinning down the density dependence of the nuclear symmetry energy has been a longstanding goal of both nuclear physics and astrophysics. Recent observations of neutron stars, in both electromagnetic and gravitational-wave spectra, have already constrained significantly the nuclear symmetry energy at high densities. The next generation of telescopes and gravitationalwave observatories will provide an unprecedented wealth of detailed observations of neutron stars which will improve further our knowledge of the density dependence of nuclear symmetry energy, and the underlying equation of state of dense neutron-rich matter. Training deep neural networks to learn a computationally efficient representation of the mapping between astrophysical observables of neutron stars, such as masses, radii, and tidal deformabilities, and the nuclear symmetry energy allows its density dependence to be determined reliably and accurately. In this work we use a deep learning approach to determine the nuclear symmetry energy as a function of density directly from observational neutron star data. We show for the first time that artificial neural networks can precisely reconstruct the nuclear symmetry energy from a set of available neutron star observables, such as, masses and radii as those measured by, e.g., the NICER mission, or masses and tidal deformabilities as measured by the LIGO/VIRGO/KAGRA gravitational-wave detectors. These results demonstrate the potential of artificial neural networks to reconstruct the symmetry energy, and the equation of state, directly from neutron star observational data, and emphasize the importance of the deep learning approach in the era of Multi-Messenger Astrophysics.

show abstract

“…价于一个矩阵，只具备线性表示能力。理论证明，只要输入层与输出层之间添加一个隐藏层，再加上非线性激活函数，原则上就可以用有限但足够数量的神经元无限逼近任意的连续实函数 [10][11] 。这个定理被称为通用近似定理 (Universal approximation theorem)，它使得我们可以针对不同的问题和数据结构，设计相应的人工神经网络，解决比如分类、回归、聚类、非线性特征降维等等复杂的问题。 Bias 输⼊层隐藏层输出层图 1 包含一个隐藏层的人工神经网络的示例通常来说，我们可以将机器学习处理问题的方法分成三类：监督学习，无监督学习和强化学习 [3][4] 1) 。它们之间的差异来自人对机器学习过程的干预程度不同：监督学习需要给定训练中输入数据的标签作为监督指导；而无监督学习是机器从数据自行挖掘其中的模式规律；强化学习则是设定反馈 (reward) 机制，通过奖励和惩罚来训练智能体 (agent) 对环境的响应。几种方式各有所长，例如监督学习在图像分类和文本理解等分类回归问题上表现出色，无监督学习适用于异常检测、推荐系统、大数据可视化等聚类降维问题，而强化学习在游戏 AI 以及天气预报等实时决策问题中大放异彩 [4][5] 。基于以上的成功应用场景，这些新型建模和数据处理方法正在物理科学研究的各个尺度上大展身手：从在数万亿天空像素中寻找系外行星，到利用大尺度结构观测推测宇宙常数，再到在大型强子对撞机事件中探测信号等 [8,[14][15] 。此外，从计算物理角度，原有的计算模拟加速效率、优化反问题求解和革新第一性原理计算技术等等目标也因机器学习社区的蓬勃发展而受益 [6,[16][17] 。除了使用机器学习方法进行科学发现和改进计算外，也有超出应用范畴对机器学习模型进行物理可解释性的探索，比如探讨深度神经网络提取数据特征和重整化方法的等价性、尝试直接引入物理规律来改进机器学习性能等 [18][19] 。以上种种应用和成果同时也在深远地影响着分子动力学、药物设计、材料设计、成像系统优化和金融系统分析等与产业结合更紧密的领域 [6,[20][21] 。在高能核物理领域，由于实验数据量大、模拟计算需求大和第一性原理计算开销大的特点，深度学习已经成为必不可少的技术。目前已在如下几个方面有广泛应用：探测器设计、粒子重建和分析、喷柱 (Jet) 分类等 [14,[22][23] ；物理模拟的加速、对物态或物理过程信息的提取等 [24][25][26][27] ；对物理表征方式的优化、第一性原理计算加速等 [28][29][30][31] ；以及对核天体物理观测的处理和物态性质的研究 [32][33][34] 。本文立足于作者最近的系列工作和新进展，管中窥豹，试图从相对论重离子碰撞和格点 QCD 计算两个方向的相关研究出发，简要阐述深度学习在关键物理信息重建、计算性能改进等方面的应用。 1) 或者按另一种分法，分成判别式学习和生成式学习，后者以表征数据背后的概率分布为核心。除此之外，还发展出了半监督学习、自监督学习、对比学习以及主动学习等弱监督式的方法 [12][13]…”

unclassified

Applications of deep learning in high energy nuclear physics

Wang¹,

Pang²,

Zhou³

2022

Sci. Sin.-Phys. Mech. Astron.

View full text Add to dashboard Cite

Extensive studies of the neutron star equation of state from the deep learning inference with the observational data augmentation

Cited by 46 publications

References 100 publications

Translating Neutron Star Observations to Nuclear Symmetry Energy via Deep Neural Networks

Translating Neutron Star Observations to Nuclear Symmetry Energy via Deep Neural Networks

Translating neutron star observations to nuclear symmetry energy via artificial neural networks

Applications of deep learning in high energy nuclear physics

Contact Info

Product

Resources

About