-
Iterated Relevance Matrix Analysis (IRMA) for the identification of class-discriminative subspaces
Authors:
Sofie Lövdal,
Michael Biehl
Abstract:
We introduce and investigate the iterated application of Generalized Matrix Learning Vector Quantizaton for the analysis of feature relevances in classification problems, as well as for the construction of class-discriminative subspaces. The suggested Iterated Relevance Matrix Analysis (IRMA) identifies a linear subspace representing the classification specific information of the considered data s…
▽ More
We introduce and investigate the iterated application of Generalized Matrix Learning Vector Quantizaton for the analysis of feature relevances in classification problems, as well as for the construction of class-discriminative subspaces. The suggested Iterated Relevance Matrix Analysis (IRMA) identifies a linear subspace representing the classification specific information of the considered data sets using Generalized Matrix Learning Vector Quantization (GMLVQ). By iteratively determining a new discriminative subspace while projecting out all previously identified ones, a combined subspace carrying all class-specific information can be found. This facilitates a detailed analysis of feature relevances, and enables improved low-dimensional representations and visualizations of labeled data sets. Additionally, the IRMA-based class-discriminative subspace can be used for dimensionality reduction and the training of robust classifiers with potentially improved performance.
△ Less
Submitted 23 January, 2024;
originally announced January 2024.
-
The Gaia DR3 view of dynamical substructure in the stellar halo near the Sun
Authors:
Emma Dodd,
Thomas M. Callingham,
Amina Helmi,
Tadafumi Matsuno,
Tomás Ruiz-Lara,
Eduardo Balbinot,
Sofie Lövdal
Abstract:
The debris from past merger events is expected and, to some extent, known to populate the stellar halo near the Sun. We aim to identify and characterise such merger debris using Gaia DR3 data supplemented by metallicity and chemical abundance information from LAMOST LRS and APOGEE for halo stars within 2.5 kpc from the Sun. We utilise a single linkage-based clustering algorithm to identify over-de…
▽ More
The debris from past merger events is expected and, to some extent, known to populate the stellar halo near the Sun. We aim to identify and characterise such merger debris using Gaia DR3 data supplemented by metallicity and chemical abundance information from LAMOST LRS and APOGEE for halo stars within 2.5 kpc from the Sun. We utilise a single linkage-based clustering algorithm to identify over-densities in Integrals of Motion space that could be due to merger debris. Combined with metallicity information and chemical abundances, we characterise these statistically significant over-densities. We find that the local stellar halo contains 7 main dynamical groups, some of in-situ and some of accreted origin, most of which are already known. We report the discovery of a new substructure, which we name ED-1. In addition, we find evidence for 11 independent smaller clumps, 5 of which are new: ED-2, 3, 4, 5 and 6 are typically rather tight dynamically, depict a small range of metallicities, and their abundances when available, as well as their location in Integrals of Motion space, suggest an accreted origin. The local halo contains an important amount of substructure, of both in-situ and accreted origin.
△ Less
Submitted 19 July, 2022; v1 submitted 22 June, 2022;
originally announced June 2022.
-
Substructure in the stellar halo near the Sun. II. Characterisation of independent structures
Authors:
Tomás Ruiz-Lara,
Tadafumi Matsuno,
S. Sofie Lövdal,
Amina Helmi,
Emma Dodd,
Helmer H. Koppelman
Abstract:
In Lövdal et al, we presented a data-driven method for clustering in Integrals of Motion space and applied it to a large sample of nearby halo stars with 6D phase-space information. We identified a large number of clusters, many of which could tentatively be merged into larger groups. Our goal is to establish the reality of the clusters through a combined study of their stellar populations to gain…
▽ More
In Lövdal et al, we presented a data-driven method for clustering in Integrals of Motion space and applied it to a large sample of nearby halo stars with 6D phase-space information. We identified a large number of clusters, many of which could tentatively be merged into larger groups. Our goal is to establish the reality of the clusters through a combined study of their stellar populations to gain more insights into the accretion history of the Milky Way. We develop a procedure that quantifies the similarity of clusters based on KS tests using their metallicity distribution functions, and an isochrone fitting method to determine their average age, which is also used to compare the distribution of stars in the Colour-Absolute magnitude diagram. This allows us to group clusters into substructures, and to compare substructures with one another. The clusters identified are merged into 12 extended substructures, while 8 small clusters remain as such. The large substructures include the previously known Gaia-Enceladus, Helmi streams, Sequoia, and Thamnos 1 and 2. We identify overdensities associated with the hot thick disc and hosting a metal-poor population. Especially notable is our largest substructure which, although peaking at the metallicity characteristic of the thick disk has a well populated metal-poor component, and dynamics in-between hot thick disc and halo. We identify additional debris in the region occupied by Sequoia with distinct kinematics, likely remnants of three different accretion events with progenitors of similar mass. We also identify different trends of [Mg/Fe] vs [Fe/H] for the various substructures confirming our dissection of the nearby halo. At least 20\% of the halo near the Sun is associated to substructures. When comparing their global properties, we note that those substructures on retrograde orbits are not only more metal-poor on average but also older.
△ Less
Submitted 28 April, 2022; v1 submitted 7 January, 2022;
originally announced January 2022.
-
Substructure in the stellar halo near the Sun. I. Data-driven clustering in Integrals of Motion space
Authors:
S. Sofie Lövdal,
Tomás Ruiz-Lara,
Helmer H. Koppelman,
Tadafumi Matsuno,
Emma Dodd,
Amina Helmi
Abstract:
Aims: Develop a data-driven and statistically based method for finding such clumps in Integrals of Motion space for nearby halo stars and evaluating their significance robustly. Methods: We use data from Gaia EDR3 extended with radial velocities from ground-based spectroscopic surveys to construct a sample of halo stars within 2.5 kpc from the Sun. We apply a hierarchical clustering method that us…
▽ More
Aims: Develop a data-driven and statistically based method for finding such clumps in Integrals of Motion space for nearby halo stars and evaluating their significance robustly. Methods: We use data from Gaia EDR3 extended with radial velocities from ground-based spectroscopic surveys to construct a sample of halo stars within 2.5 kpc from the Sun. We apply a hierarchical clustering method that uses the single linkage algorithm in a 3D space defined by the commonly used integrals of motion energy $E$, together with two components of the angular momentum, $L_z$ and $L_\perp$. To evaluate the statistical significance of the clusters found, we compare the density within an ellipsoidal region centered on the cluster to that of random sets with similar global dynamical properties. We pick out the signal at the location of their maximum statistical significance in the hierarchical tree. We estimate the proximity of a star to the cluster center using the Mahalanobis distance. We also apply the HDBSCAN clustering algorithm in velocity space. Results: Our procedure identifies 67 highly significant clusters ($ > 3σ$), containing 12\% of the sources in our halo set, and in total 232 subgroups or individual streams in velocity space. In total, 13.8\% of the stars in our data set can be confidently associated to a significant cluster based on their Mahalanobis distance. Inspection of our data set reveals a complex web of relationships between the significant clusters, suggesting that they can be tentatively grouped into at least 6 main structures, many of which can be associated to previously identified halo substructures, and a number of independent substructures. This preliminary conclusion is further explored in an accompanying paper by Ruiz-Lara et al., where we also characterize the substructures in terms of their stellar populations. Conclusions: We find... (abridged version)
△ Less
Submitted 17 May, 2022; v1 submitted 7 January, 2022;
originally announced January 2022.
