-
The Three Hundred project: Estimating the dependence of gas filaments on the mass of galaxy clusters
Authors:
Sara Santoni,
Marco De Petris,
Gustavo Yepes,
Antonio Ferragamo,
Matteo Bianconi,
Meghan E. Gray,
Ulrike Kuchner,
Frazer R. Pearce,
Weiguang Cui,
Stefano Ettori
Abstract:
Galaxy clusters are located in the densest areas of the universe and are intricately connected to larger structures through the filamentary network of the Cosmic Web. In this scenario, matter flows from areas of lower density to higher density. As a result, the properties of galaxy clusters are deeply influenced by the filaments that are attached to them, which are quantified by a parameter known…
▽ More
Galaxy clusters are located in the densest areas of the universe and are intricately connected to larger structures through the filamentary network of the Cosmic Web. In this scenario, matter flows from areas of lower density to higher density. As a result, the properties of galaxy clusters are deeply influenced by the filaments that are attached to them, which are quantified by a parameter known as connectivity. We explore the dependence of gas-traced filaments connected to galaxy clusters on the mass and dynamical state of the cluster. Moreover, we evaluate the effectiveness of the cosmic web extraction procedure from the gas density maps of simulated cluster regions. Using the DisPerSE cosmic web finder, we identify filamentary structures from 3D gas particle distribution in 324 simulated regions of $30 \, h^{-1}$ Mpc side from The Three Hundred hydrodynamical simulation at redshifts z=0, 1, and 2. We estimate the connectivity at various apertures for $\sim3000$ groups and clusters spanning a mass range from $10^{13} \, h^{-1} \, M_{\odot}$ to $10^{15} \, h^{-1} \, M_{\odot}$. Relationships between connectivity and cluster properties like radius, mass, dynamical state and hydrostatic mass bias are explored. We show that the connectivity is strongly correlated with the mass of galaxy clusters, with more massive clusters being on average more connected. This finding aligns with previous studies in literature, both from observational and simulated data sets. Additionally, we observe a dependence of the connectivity on the aperture at which it is estimated. We find that connectivity decreases with cosmic time, while no dependencies on the dynamical state and hydrostatic mass bias of the cluster are found. Lastly, we observe a significant agreement between the connectivity measured from gas-traced and mock-galaxies-traced filaments in the simulation.
△ Less
Submitted 27 May, 2024;
originally announced May 2024.
-
Impact of filaments on galaxy cluster properties in The Three Hundred simulation
Authors:
Sara Santoni,
Marco De Petris,
Antonio Ferragamo,
Gustavo Yepes,
Weiguang Cui
Abstract:
Galaxy clusters and their filamentary outskirts reveal useful laboratories to test cosmological models and investigate Universe composition and evolution. Their environment, in particular the filaments of the Cosmic Web to which they are connected, plays an important role in sha** the properties of galaxy clusters. In this project, we analyse the gas filamentary structures present in 324 regions…
▽ More
Galaxy clusters and their filamentary outskirts reveal useful laboratories to test cosmological models and investigate Universe composition and evolution. Their environment, in particular the filaments of the Cosmic Web to which they are connected, plays an important role in sha** the properties of galaxy clusters. In this project, we analyse the gas filamentary structures present in 324 regions of The Three Hundred hydrodynamical simulation extracted with the DisPerSE filament finder. We estimate the number of gas filaments globally connected to several galaxy clusters, i.e. the connectivity k, with a mass range of $10^{13} \leq M_{200} \, h^{-1} \, M_{\odot} \leq 10^{15} $ at redshift $z=0$. We study the positive correlation between the connectivity and mass of galaxy clusters. Moreover, we explore the impact of filaments on the dynamical state of clusters, quantified by the degree of relaxation parameter $χ$.
△ Less
Submitted 11 October, 2023;
originally announced October 2023.
-
Robust Classification via Support Vector Machines
Authors:
Vali Asimit,
Ioannis Kyriakou,
Simone Santoni,
Salvatore Scognamiglio,
Rui Zhu
Abstract:
Classification models are very sensitive to data uncertainty, and finding robust classifiers that are less sensitive to data uncertainty has raised great interest in the machine learning literature. This paper aims to construct robust \emph{Support Vector Machine} classifiers under feature data uncertainty via two probabilistic arguments. The first classifier, \emph{Single Perturbation}, reduces t…
▽ More
Classification models are very sensitive to data uncertainty, and finding robust classifiers that are less sensitive to data uncertainty has raised great interest in the machine learning literature. This paper aims to construct robust \emph{Support Vector Machine} classifiers under feature data uncertainty via two probabilistic arguments. The first classifier, \emph{Single Perturbation}, reduces the local effect of data uncertainty with respect to one given feature and acts as a local test that could confirm or refute the presence of significant data uncertainty for that particular feature. The second classifier, \emph{Extreme Empirical Loss}, aims to reduce the aggregate effect of data uncertainty with respect to all features, which is possible via a trade-off between the number of prediction model violations and the size of these violations. Both methodologies are computationally efficient and our extensive numerical investigation highlights the advantages and possible limitations of the two robust classifiers on synthetic and real-life data.
△ Less
Submitted 28 February, 2022; v1 submitted 27 April, 2021;
originally announced April 2021.
