arXiv subjects

Kohei Hayashi

Publications and source records attributed to Kohei Hayashi.

At least 91 records · Page 5Linked to original sources

arXiv2019

Data Interpolating Prediction: Alternative Interpretation of Mixup

Data augmentation by mixing samples, such as Mixup, has widely been used typically for classification tasks. However, this strategy is not always effective due to the gap between augmented samples for training and original samples for testing. This gap may prevent a classifier from learning the optimal decision boundary and increase the generalization error. To overcome this problem, we propose an alternative framework called Data Interpolating Prediction (DIP). Unlike common data augmentations, we encapsulate the sample-mixing process in the hypothesis class of a classifier so that train and test samples are treated equally. We derive the generalization bound and show that DIP helps to reduce the original Rademacher complexity. Also, we empirically demonstrate that DIP can outperform existing Mixup.

cs.LG↗

arXiv2019

Boötes IV: A New Milky Way Satellite Discovered in the Subaru Hyper Suprime-Cam Survey and Implications for the Missing Satellite Problem

We report on the discovery of a new Milky Way (MW) satellite in Boötes based on data from the on-going Hyper Suprime-Cam (HSC) Subaru Strategic Program (SSP). This satellite, named Boötes IV, is the third ultra-faint dwarf that we have discovered in the HSC-SSP. We have identified a statistically significant (32.3$σ$) overdensity of stars having characteristics of a metal-poor, old stellar population. The distance to this stellar system is $D_{\odot}=209^{+20}_{-18}$ kpc with a $V$-band absolute magnitude of $M_V=-4.53^{+0.23}_{-0.21}$ mag. Boötes IV has a half-light radius of $r_h=462^{+98}_{-84}$ pc and an ellipticity of $0.64^{+0.05}_{-0.05}$, which clearly suggests that this is a dwarf satellite galaxy. We also found another overdensity that appears to be a faint globular cluster with $M_V=-0.20^{+0.59}_{-0.83}$ mag and $r_h=5.9^{+1.5}_{-1.3}$ pc located at $D_{\odot}=46^{+4}_{-4}$ kpc. Adopting the recent prediction for the total population of satellites in a MW-sized halo by Newton et al. (2018), which combined the characteristics of the observed satellites by SDSS and DES with the subhalos obtained in $Λ$CDM models, we estimate that there should be about two MW satellites at $M_V\le0$ in the $\sim676$ deg$^2$ covered by HSC-SSP, whereas that area includes six satellites. Thus, the observed number of satellites is larger than the theoretical prediction. On the face of it, we have a problem of too many satellites, instead of the well-known missing satellites problem whereby the $Λ$CDM theory overpredicts the number of satellites in a MW-sized halo. This may imply that the models need more refinements for the assignment of subhalos to satellites such as considering those found by the current deeper survey. [abridged]

astro-ph.GA↗

arXiv2019

The stellar halo of the Milky Way traced by blue horizontal-branch stars in the Subaru Hyper Suprime-Cam Survey

We report on the global structure of the Milky Way (MW) stellar halo up to its outer boundary based on the analysis of blue-horizontal branch stars (BHBs). These halo tracers are extracted from the $(g,r,i,z)$ band multi-photometry in the internal data release of the on-going Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) surveyed over $\sim550$~deg$^2$ area. In order to select most likely BHBs by removing blue straggler stars (BSs) and other contamination in a statistically significant manner, we have developed and applied an extensive Bayesian method, instead of the simple color cuts adopted in our previous work, where each of the template BHBs and non-BHBs obtained from the available catalogs is represented as a mixture of multiple Gaussian distributions in the color-color diagrams. We found from the candidate BHBs in the range of 18.5 r_{\rm b}$. The latter profile with a prolate shape having an axial ratio of $q=1.72^{+0.44}_{-0.28}$ is most likely and this halo may hold a rather sharp boundary at r=160kpc. The slopes of the halo density profiles are compared with those from the suite of hydrodynamical simulations for the formation of stellar halos. This comparison suggests that the MW stellar halo may consist of the two overlapping components: the in situ. inner halo as probed by RR Lyrae stars showing a relatively steep radial density profile and the ex situ. outer halo with a shallow profile probed by BHBs here, which is made by accretion of small stellar systems.

astro-ph.GA↗

arXiv2019

Imaging the Inner Structure of a Nuclear Reactor by Cosmic Muon Radiography

We studied the inner structure of the nuclear reactor of the Japan Atomic Power Company (JAPC) at Tokai, Japan, by the muon radiography. In this study, muon detectors were placed outside of the reactor building. By detecting cosmic muons penetrating through the wall of the reactor building, we could successfully identify the objects such as the containment vessel, pressure vessel, and other structures of the reactor. We also observed a concentration of heavy material which can be attributed to the nuclear fuel assemblies stored in the nuclear fuel storage pool.

physics.ins-det↗

arXiv2019

On Random Subsampling of Gaussian Process Regression: A Graphon-Based Analysis

In this paper, we study random subsampling of Gaussian process regression, one of the simplest approximation baselines, from a theoretical perspective. Although subsampling discards a large part of training data, we show provable guarantees on the accuracy of the predictive mean/variance and its generalization ability. For analysis, we consider embedding kernel matrices into graphons, which encapsulate the difference of the sample size and enables us to evaluate the approximation and generalization errors in a unified manner. The experimental results show that the subsampling approximation achieves a better trade-off regarding accuracy and runtime than the Nyström and random Fourier expansion methods.

stat.ML↗

arXiv2018

Effects of mass models on dynamical mass estimate: the case of ultra diffuse galaxy NGC1052-DF2

NGC1052-DF2 was recently discovered as the dark-matter deficient galaxy claimed by van Dokkum et al. (2018a, vD18). However, large uncertainties on its dynamical mass estimate have been pointed out, concerning the paucity of sample, statistical methods and distance measurements. In this work, we discuss the effects of the difference in modeling of the tracer profile of this galaxy on the dynamical mass estimate. To do this, we assume that the tracer densities are modeled with power-law and Sérsic profiles, and then we solve the spherical Jeans equation to estimate the dynamical mass. Applying these models to kinematic data of globular clusters in NGC1052-DF2, we compare 90 per cent upper limits of dynamical mass-to-light ratios estimated between from this analysis and from vD18. We find that the upper limit obtained by the power-law is virtually the same as the result from vD18, whilst this limit estimated by the Sérsic is significantly greater than that from vD18, thereby suggesting that NGC1052-DF2 can still be a dark-matter dominated system. Consequently, we propose that dynamical mass estimate of a galaxy is largely affected by not only small kinematic sample but the choice of tracer distributions, and thus the estimated mass still remains quite uncertain.

astro-ph.GA↗

arXiv2018

Foreground effect on the $J$-factor estimation of ultra-faint dwarf spheroidal galaxies

Dwarf spheroidal galaxies (dSphs) are promising targets for the gamma-ray dark matter (DM) search. In particular, DM annihilation signal is expected to be strong in some of the recently discovered nearby ultra-faint dSphs, which potentially give stringent constraints on the $\mathcal{O}(1)$ TeV WIMP DM. However, various non-negligible systematic uncertainties complicate the estimation of the astrophysical factors relevant for the DM search in these objects. Among them, the effects of foreground stars particularly attract attention because the contamination is unavoidable even for the future kinematical survey. In this article, we assess the effects of the foreground contamination on the astrophysical $J$-factor estimation by generating mock samples of stars in the four ultra-faint dSphs and using a model of future spectrographs. We investigate various data cuts to optimize the quality of the data and apply a likelihood analysis which takes member and foreground stellar distributions into account. We show that the foreground star contaminations in the signal region (the region of interest) and their statistical uncertainty can be estimated by interpolating the foreground star distribution in the control region where the foreground stars dominate the member stars. Such regions can be secured at future spectroscopic observations utilizing a multiple object spectrograph with a large field of view; e.g. the Prime Focus Spectrograph mounted on Subaru Telescope. The above estimation has several advantages: The data-driven estimation of the contamination makes the analysis of the astrophysical factor stable against the complicated foreground distribution. Besides, foreground contamination effect is considered in the likelihood analysis.

astro-ph.GA↗

arXiv2018

Dark halo structure in the Carina dwarf spheroidal galaxy: joint analysis of multiple stellar components

Photometric and spectroscopic observations of the Carina dSph revealed that this galaxy contains two dominant stellar populations of different age and kinematics. The co-existence of multiple populations provides new constraints on the dark halo structure of the galaxy, because different populations should be in equilibrium in the same dark matter potential well. We develop non-spherical dynamical models including such multiple stellar components and attempt to constrain the properties of the non-spherical dark halo of Carina. We find that Carina probably has a larger and denser dark halo than found in previous works and a less cuspy inner dark matter density profile, even though the uncertainties of dark halo parameters are still large due to small volume of data sample. Using our fitting results, we evaluate astrophysical factors for dark matter annihilation and decay and find that Carina should be one of the most promising detectable targets among classical dSph galaxies. We also calculate stellar velocity anisotropy profiles for both stellar populations and find that they are both radially anisotropic in the inner regions, while in the outer regions the older population becomes more tangentially biased than the intermediate one. This is consistent with the anisotropy predicted from tidal effects on the dynamical structure of a satellite galaxy and thereby can be considered as kinematic evidence for the tidal evolution of Carina.

astro-ph.GA↗

arXiv2018

The Missing Satellite Problem Outside of the Local Group: I -- Pilot Observation

We present results from a pilot observation of nearby (~20 Mpc) galaxies with mass similar to that of the Milky Way (MW) to address the missing satellite problem. This is the first paper from an on-going project to address the problem with a statistical sample of galaxies outside of the Local Group (LG) without employing an assumption that the LG is a typical halo in the Universe. Thanks to the close distances of our targets, dwarf galaxies around them can be identified as extended, diffuse galaxies. By applying a surface brightness cut together with a careful visual screening to remove artifacts and background contamination, we construct a sample of dwarf galaxies. The luminosity function (LF) of one of the targets is broadly consistent with that of the MW, but the other has a more abundant dwarf population. Numerical simulations by Okamoto (2013) seem to overpredict the number of dwarfs on average, while more recent predictions from Copernicus Complexio are in a better agreement. In both observations and simulations, there is a large diversity in the LFs, demonstrating the importance of addressing the missing satellite problem with a statistically representative sample. We also characterize the projected spatial distributions of the satellites and do not observe strong evidence for alignments around the central galaxies. Based on this successful pilot observation, we are carrying out further observations to increase the sample of nearby galaxies, which we plan to report in our future paper.

astro-ph.GA↗

arXiv2017

Stellar Stream and Halo Structure in the Andromeda Galaxy From a Subaru/Hyper Suprime-Cam Survey

We present wide and deep photometry of the northwest part of the halo of the Andromeda galaxy (M31) using Hyper Suprime-Cam on the Subaru Telescope. The survey covers 9.2 deg$^{2}$ field in the $g$, $i$, and $NB515$ bands and shows a clear red giant branch (RGB) of M31's halo stars and a pronounced red clump (RC) feature. The spatial distribution of RC stars shows a prominent stream feature, the North Western (NW) Stream, and a diffuse substructure in the south part of our survey field. We estimate the distances based on the RC method and obtain $(m-M)$ = 24.63$\pm 0.191$(random)$\pm0.057$(systematic) and 24.29$\pm 0.211$(random)$\pm0.057$(systematic) mag for the NW stream and diffuse substructure, respectively, implying that the NW Stream is located behind M31, whereas the diffuse substructure is located in front. We also estimate line-of-sight distances along the NW Stream and find that the south part of the stream is $\sim$20 kpc closer to us relative to the north part. The distance to the NW Stream inferred from the isochrone fitting to the color-magnitude diagram favors the RC-based distance, but the TRGB-based distance estimated for $NB515$-selected RGB stars does not agree with it. The surface number density distribution of RC stars across the NW Stream is found to be approximately Gaussian with a FWHM of $\sim$25 arcmin (5.7 kpc), with a slight skew to the south-west side. That along the NW Stream shows a complicated structure including variations in number density and a significant gap in the stream.

astro-ph.GA↗

arXiv2017

Discovery of a big void in Khufu's Pyramid by observation of cosmic-ray muons

The Great Pyramid or Khufu's Pyramid was built on the Giza Plateau (Egypt) during the IVth dynasty by the pharaoh Khufu (Cheops), who reigned from 2509 to 2483 BC. Despite being one of the oldest and largest monuments on Earth, there is no consensus about how it was built. To better understand its internal structure, we imaged the pyramid using muons, which are by-products of cosmic rays that are only partially absorbed by stone. The resulting cosmic-ray muon radiography allows us to visualize the known and potentially unknown voids in the pyramid in a non-invasive way. Here we report the discovery of a large void (with a cross section similar to the Grand Gallery and a length of 30 m minimum) above the Grand Gallery, which constitutes the first major inner structure found in the Great Pyramid since the 19th century. This void, named ScanPyramids Big Void, was first observed with nuclear emulsion films installed in the Queen's chamber (University of Nagoya), then confirmed with scintillator hodoscopes set up in the same chamber (KEK) and re-confirmed with gas detectors outside of the pyramid (CEA). This large void has therefore been detected with a high confidence by three different muon detection technologies and three independent analyses. These results constitute a breakthrough for the understanding of Khufu's Pyramid and its internal structure. While there is currently no information about the role of this void, these findings show how modern particle physics can shed new light on the world's archaeological heritage.

physics.ins-det↗

arXiv2017

Think Globally, Embed Locally --- Locally Linear Meta-embedding of Words

Distributed word embeddings have shown superior performances in numerous Natural Language Processing (NLP) tasks. However, their performances vary significantly across different tasks, implying that the word embeddings learnt by those methods capture complementary aspects of lexical semantics. Therefore, we believe that it is important to combine the existing word embeddings to produce more accurate and complete \emph{meta-embeddings} of words. For this purpose, we propose an unsupervised locally linear meta-embedding learning method that takes pre-trained word embeddings as the input, and produces more accurate meta embeddings. Unlike previously proposed meta-embedding learning methods that learn a global projection over all words in a vocabulary, our proposed method is sensitive to the differences in local neighbourhoods of the individual source word embeddings. Moreover, we show that vector concatenation, a previously proposed highly competitive baseline approach for integrating word embeddings, can be derived as a special case of the proposed method. Experimental results on semantic similarity, word analogy, relation classification, and short-text classification tasks show that our meta-embeddings to significantly outperform prior methods in several benchmark datasets, establishing a new state of the art for meta-embeddings.

cs.CL↗

arXiv2017

On Tensor Train Rank Minimization: Statistical Efficiency and Scalable Algorithm

Tensor train (TT) decomposition provides a space-efficient representation for higher-order tensors. Despite its advantage, we face two crucial limitations when we apply the TT decomposition to machine learning problems: the lack of statistical theory and of scalable algorithms. In this paper, we address the limitations. First, we introduce a convex relaxation of the TT decomposition problem and derive its error bound for the tensor completion task. Next, we develop an alternating optimization method with a randomization technique, in which the time complexity is as efficient as the space complexity is. In experiments, we numerically confirm the derived bounds and empirically demonstrate the performance of our method with a real higher-order tensor.

stat.ML↗

arXiv2017

Impact of axisymmetric mass models for dwarf spheroidal galaxies on indirect dark matter searches

Dwarf spheroidals are low-luminosity satellite galaxies of the Milky Way highly dominated by dark matter (DM). Therefore, they are prime targets to search for signals from dark matter annihilation using gamma-ray observations. While the typical assumption is that the dark matter density profile of these satellite galaxies can be described by a spherical symmetric Navarro-Frenk-White (NFW) profile, recent observational data of stellar kinematics suggest that the DM halos around these galaxies are better described by axisymmetric profiles. Motivated by such evidence, we analyse about seven years of PASS8 Fermi data for seven classical dwarf galaxies, including Draco, adopting both the widely used NFW profile and observationally-motivated axisymmetric density profiles. For four of the selected dwarfs (Sextans, Carina, Sculptor and Fornax) axisymmetric mass models suggest a cored density profile rather than the commonly adopted cusped profile. We found that upper limits on the annihilation cross section for some of these dwarfs are significantly higher than the ones achieved using an NFW profile. Therefore, upper limits in the literature obtained using spherical symmetric cusped profiles, such as the NFW, might be overestimated. Our results show that it is extremely important to use observationally motivated density profiles going beyond the usually adopted NFW in order to obtain accurate constraints on the dark matter annihilation cross section.

astro-ph.CO↗

arXiv2017

Searches for New Milky Way Satellites from the First Two Years of Data of the Subaru/Hyper Suprime-Cam Survey: Discovery of Cetus~III

We present the results from a search for new Milky Way (MW) satellites from the first two years of data from the Hyper Suprime-Cam (HSC) Subaru Strategic Program (SSP) $\sim 300$~deg$^2$ and report the discovery of a highly compelling ultra-faint dwarf galaxy candidate in Cetus. This is the second ultra-faint dwarf we have discovered after Virgo~I reported in our previous paper. This satellite, Cetus~III, has been identified as a statistically significant (10.7$σ$) spatial overdensity of star-like objects, which are selected from a relevant isochrone filter designed for a metal-poor and old stellar population. This stellar system is located at a heliocentric distance of 251$^{+24}_{-11}$~kpc with a most likely absolute magnitude of $M_V = -2.4 \pm 0.6$~mag estimated from a Monte Carlo analysis. Cetus~III is extended with a half-light radius of $r_h = 90^{+42}_{-17}$~pc, suggesting that this is a faint dwarf satellite in the MW located beyond the detection limit of the Sloan Digital Sky Survey. Further spectroscopic studies are needed to assess the nature of this stellar system. We also revisit and update the parameters for Virgo~I finding $M_V = -0.33^{+0.75}_{-0.87}$~mag and $r_h = 47^{+19}_{-13}$~pc. Using simulations of $Λ$-dominated cold dark matter models, we predict that we should find one or two new MW satellites from $\sim 300$~deg$^2$ HSC-SSP data, in rough agreement with the discovery rate so far. The further survey and completion of HSC-SSP over $\sim 1,400$~deg$^2$ will provide robust insights into the missing satellites problem.

astro-ph.GA↗

arXiv2017

Foreground effect on the $J$-factor estimation of classical dwarf spheroidal galaxies

The gamma-ray observation of the dwarf spheroidal galaxies (dSphs) is a promising approach to search for the dark matter annihilation (or decay) signal. The dSphs are the nearby satellite galaxies with a clean environment and dense dark matter halo so that they give stringent constraints on the 1 TeV dark matter. However, recent studies have revealed that current estimation of astrophysical factors relevant for the dark matter searches are not conservative, where the various non-negligible systematic uncertainties are not taken into account. Among them, the effect of foreground stars on the astrophysical factors has not been paid much attention, which becomes more important for deeper and wider stellar surveys in the future. In this article, we assess the effects of the foreground contamination by generating the mock samples of stars and using a model of future spectrographs. We investigate various data cuts to optimize the quality of the data and find that the cuts on the velocity and surface gravity can efficiently eliminate the contamination. We also propose a new likelihood function which includes the foreground distribution function. We apply this likelihood function to the fit of the three types of the mock data (Ursa Minor, Draco with large dark matter halo, and Draco with small halo) and three cases of the observation. The likelihood successfully reproduces the input $J$-factor value while the fit without considering the foreground distribution gives large deviation from the input value by a factor of three.

astro-ph.GA↗

arXiv2017

Universal dark halo scaling relation for the dwarf spheroidal satellites

Motivated by a recently found interesting property of the dark halo surface density within a radius, $r_{\rm max}$, giving the maximum circular velocity, $V_{\rm max}$, we investigate it for dark halos of the Milky Way's and Andromeda's dwarf satellites based on cosmological simulations. We select and analyze the simulated subhalos associated with Milky Way-sized dark halos and find that the values of their surface densities, $Σ_{V_{\rm max}}$, are in good agreement with those for the observed dwarf spheroidal satellites even without employing any fitting procedures. This implies that this surface density would not be largely affected by any baryonic feedbacks and thus universal. Moreover, all subhalos on the small scales of dwarf satellites are expected to obey the relation $Σ_{V_{\rm max}}\propto V_{\rm max}$, irrespective of differences in their orbital evolutions, host halo properties, and observed redshifts. Therefore, we find that the universal scaling relation for dark halos on dwarf galaxy mass scales surely exists and provides us important clues to understanding fundamental properties of dark halos. We also investigate orbital and dynamical evolutions of subhalos to understand the origin of this universal dark halo relation and find that most of subhalos evolve generally along the $r_{\rm max}\propto V_{\rm max}$ sequence, even though these subhalos have undergone different histories of mass assembly and tidal stripping. This sequence, therefore, should be the key feature to understand the nature of the universality of $Σ_{V_{\rm max}}$.

astro-ph.GA↗

arXiv2017

Making Tree Ensembles Interpretable: A Bayesian Model Selection Approach

Tree ensembles, such as random forests and boosted trees, are renowned for their high prediction performance. However, their interpretability is critically limited due to the enormous complexity. In this study, we present a method to make a complex tree ensemble interpretable by simplifying the model. Specifically, we formalize the simplification of tree ensembles as a model selection problem. Given a complex tree ensemble, we aim at obtaining the simplest representation that is essentially equivalent to the original one. To this end, we derive a Bayesian model selection algorithm that optimizes the simplified model while maintaining the prediction performance. Our numerical experiments on several datasets showed that complicated tree ensembles were reasonably approximated as interpretable.

stat.ML↗