The ability to obtain reliable point estimates of model parameters is of crucial importance in many fields of physics. This is often a difficult task given that the observed data can have a very high number of dimensions. In order to address this problem, we propose a novel approach to construct parameter estimators with a quantifiable bias using an order expansion of highly compressed deep summary statistics of the observed data. These summary statistics are learned automatically using an information maximising loss. Given an observation, we further show how one can use the constructed estimators to obtain approximate Bayes computation (ABC) posterior estimates and their corresponding uncertainties that can be used for parameter inference using Gaussian process regression even if the likelihood is not tractable. We validate our method with an application to the problem of cosmological parameter inference of weak lensing mass maps. We show in that case that the constructed estimators are unbiased and have an almost optimal variance, while the posterior distribution obtained with the Gaussian process regression is close to the true posterior and performs better or equally well than comparable methods.
Low-cost neural summaries for cosmological inference: an ensemble of convolutional neural networks trained on approximate simulations
Università degli Studi di Milano(米兰大学);INFN, Sezione di Milano(意大利国家核物理研究所米兰分部);INAF, Osservatorio Astronomico di Brera(意大利国家天体物理研究所布雷拉天文台);Astronomical Observatory of the Autonomous Region of Aosta Valley (OAVdA), Fondazione Clément Fillietroz(瓦莱达奥斯塔大区自治领天文台(克莱门特·菲莱特罗兹基金会))
From ANN to BNN: Inferring Reionization Parameters using Uncertainty-aware Emulators of 21-cm Summaries
Indian Institute of Technology Indore(印度印多尔理工学院);The Open University of Israel(以色列开放大学);University of Cambridge(剑桥大学);University of Manchester(曼彻斯特大学);Curtin University(科廷大学)
Hybrid summary statistics: neural weak lensing inference beyond the power spectrum
Imperial Centre for Inference and Cosmology (ICIC)(帝国推理与宇宙学中心);Imperial Astrophysics(帝国天文学);Department of Physics(物理系);Sorbonne Université(索邦大学);CNRS(法国国家科学研究中心);Institut d’Astrophysique de Paris(巴黎天体物理研究所);Center for Computational Astrophysics(计算天文学中心);Flatiron Institute(Flatiron研究院)
Flinch: A Differentiable Framework for Field-Level Inference of Cosmological parameters from curved sky data
Waterloo Centre for Astrophysics, University of Waterloo(滑铁卢天体物理中心,滑铁卢大学);Department of Physics and Astronomy, University of Waterloo(物理学与天文学系,滑铁卢大学);Oskar Klein Centre for Cosmoparticle Physics, Department of Physics, Stockholm University(奥斯卡·克莱因宇宙粒子物理中心,物理学系,斯德哥尔摩大学);Astrophysics Group, Blackett Laboratory, Imperial College London(天体物理学组,布莱克特实验室,伦敦帝国理工学院);Department of Physics and Astronomy, University College London(物理学与天文学系,伦敦大学学院);Advanced Research Computing Centre, University College London(高级研究计算中心,伦敦大学学院);Department of Physics, Royal Holloway, University of London(物理学系,皇家霍洛威学院,伦敦大学);Department of Physics, University of California, Berkeley(物理学系,加州大学伯克利分校);Berkeley Center for Cosmological Physics, UC Berkeley(伯克利宇宙物理学中心,加州大学伯克利分校);Center for Computational Astrophysics, Flatiron Institute(计算天体物理学中心,Flatiron研究所);Department of Astrophysical Sciences, Princeton University(天体物理科学系,普林斯顿大学);Department of Physics, University of California, Davis(物理学系,加州大学戴维斯分校);Lawrence Berkeley National Laboratory(劳伦斯伯克利国家实验室)
C3NN-SBI: Learning Hierarchies of $N$-Point Statistics from Cosmological Fields with Physics-Informed Neural Networks
Ludwig-Maximilians-Universität(慕尼黑大学);Flatiron Institute(平顿研究所);Excellence Cluster ORIGINS(ORIGINS卓越集群);Max-Planck-Institut für extraterrestrische Physik(马克斯·普朗克地外物理研究所);University of Arizona(亚利桑那大学);Universität Bielefeld(比勒费尔德大学)
Zel'dovich smearing approximation of the BAO feature for model-agnostic cosmological inference
Inter-University Centre for Astronomy & Astrophysics(星际大学天文学与天体物理中心);Center for Particle Cosmology, University of Pennsylvania(宾夕法尼亚大学粒子宇宙学中心);The Abdus Salam International Center for Theoretical Physics(阿卜杜斯·萨拉姆国际理论物理中心)
Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI
Kavli IPMU (WPI), UTIAS, The University of Tokyo, Japan(Kavli IPMU(WPI)、UTIAS、东京大学);Center for Data-Driven Discovery, Kavli IPMU (WPI), Japan(数据驱动发现中心、Kavli IPMU(WPI)、日本);Center for Computational Astrophysics, Flatiron Institute, New York, USA(计算天体物理学中心、Flatiron研究所、美国纽约);Department of Astrophysical Sciences, Princeton University, USA(天体物理学系、普林斯顿大学、美国);RCAST, The University of Tokyo, Japan(RCAST、东京大学、日本)
Field-level multi-tracers simulation-based inference of cosmological parameters from 3D maps
SISSA(国际高等研究学院);INFN Sezione di Trieste(意大利国家核物理研究所的里雅斯特分部);ICSC – Centro Nazionale di Ricerca in High Performance Computing, Big Data e Quantum Computing(高性能计算、大数据与量子计算国家研究中心);Imperial College London(帝国理工学院);IFPU - Institute for Fundamental Physics of the Universe(宇宙基础物理研究所)
Cosmological Field Emulation and Parameter Inference with Diffusion Models
Harvard University(哈佛大学);The NSF AI Institute for Artificial Intelligence and Fundamental Interactions(NSF人工智能与基础相互作用人工智能研究所);Massachusetts Institute of Technology(麻省理工学院)
Bayesian cosmological inference through implicit cross-correlation statistics
对下一代星系数据的分析需要准确处理系统效应,如观测星系与底层物质密度场之间的偏差。然而,所提出的模型的现象是数值昂贵或太不准确,以实现无偏的宇宙学参数的推断,即使在轻度非线性尺度的数据。作为一种替代构建准确的星系偏差模型,需要了解星系形成,我们建议构建似然分布的贝叶斯正向建模方法是不敏感的线性,尺度依赖的偏见,并提供对模型误指定的鲁棒性。我们使用最大熵参数来构造仅考虑数据和推断量之间的相关性的似然分布。通过设计,这些相关性对线性星系偏置关系不敏感,提供了所需的鲁棒性。该方法的实施和测试中的马尔可夫链蒙特卡罗方法。该方法是评估使用晕模拟目录的基础上标准的完整,宇宙学,N体模拟。我们得到公正的和严格的约束宇宙学参数只利用线性互相关率为$k\le 0.10$兆秒钟/小时。对质量约为10 $^{12}$ M $_\odot$到约10 $^{13}$ M $_\odot$的晕的测试表明,在获得宇宙学的鲁棒约束时,可以忽略线性的、尺度相关的、偏置函数的所有细节。我们的研究结果提供了一个很有前途的路径向前分析星系调查,而不需要准确地模拟星系偏置的细节,但通过设计强大的似然推理。
Differentiable Cosmological Hydrodynamics for Field-Level Inference and High Dimensional Parameter Constraints
Kavli IPMU (WPI), UTIAS, The University of Tokyo(东京大学 Kavli宇宙物理数学研究所(WPI)UTIAS);Center for Data-Driven Discovery, Kavli IPMU (WPI), UTIAS, The University of Tokyo(东京大学 Kavli宇宙物理数学研究所(WPI)UTIAS 数据驱动发现中心);Lawrence Berkeley National Lab(劳伦斯伯克利国家实验室)
Impact of non-Gaussian likelihood on cosmological constraints from the thermal Sunyaev--Zel'dovich power spectrum: a simulation-based inference analysis
Institute of Astronomy, University of Cambridge(剑桥大学天文学研究所);Kavli Institute for Cosmology, University of Cambridge(剑桥大学卡弗里宇宙学研究所);Cavendish Astrophysics, University of Cambridge(剑桥大学卡文迪许天体物理学)