Most of machine learning approaches have stemmed from the application of
minimizing the mean squared distance principle, based on the computationally
efficient quadratic optimization methods. However, when faced with
high-dimensional and noisy data, the quadratic error functionals demonstrated
many weaknesses including high sensitivity to contaminating factors and
dimensionality curse. Therefore, a lot of recent applications in machine
learning exploited properties of non-quadratic error functionals based on $L_1$
norm or even sub-linear potentials corresponding to quasinorms $L_p$ ($0<p<1$).
The back side of these approaches is increase in computational cost for
optimization. Till so far, no approaches have been suggested to deal with {\it
arbitrary} error functionals, in a flexible and computationally efficient
framework. In this paper, we develop a theory and basic universal data
approximation algorithms ($k$-means, principal components, principal manifolds
and graphs, regularized and sparse regression), based on piece-wise quadratic
error potentials of subquadratic growth (PQSQ potentials). We develop a new and
universal framework to minimize {\it arbitrary sub-quadratic error potentials}
using an algorithm with guaranteed fast convergence to the local or global
error minimum. The theory of PQSQ potentials is based on the notion of the cone
of minorant functions, and represents a natural approximation formalism based
on the application of min-plus algebra. The approach can be applied in most of
existing machine learning methods, including methods of data approximation and
regularized and sparse regression, leading to the improvement in the
computational cost/accuracy trade-off. We demonstrate that on synthetic and
real-life datasets PQSQ-based machine learning methods achieve orders of
magnitude faster computational performance than the corresponding
state-of-the-art methods.
Authors' comments: Edited and extended version with algortihms of regularized regression
We report the detection of a significant infrared variability of the nearest
tidal disruption event (TDE) ASASSN-14li using Wide-field Infrared Survey
Explorer} and newly released Near-Earth Object WISE Reactivation} data. In
comparison with the quiescent state, the infrared flux is brightened by 0.12
and 0.16 magnitude in the W1 ($3.4\mu$m) and W2 ($4.6\mu$m) bands at 36 days
after the optical discovery (or $\sim110$ days after the peak disruption date).
The flux excess is still detectable $\sim170$ more days later. Assuming that
the flare-like infrared emission is from the dust around the black hole, its
blackbody temperature is estimated to be $\sim2.1\times10^3$~K, slightly higher
than the dust sublimation temperature, indicating that the dust is likely
located close to the dust sublimation radius. The equilibrium between the
heating and radiation of the dust claims a bolometric luminosity of
$\sim10^{43}-10^{45}$~\lum, comparable with the observed peak luminosity. This
result has for the first time confirmed the detection of infrared emission from
the dust echoes of TDEs.
Authors' comments: published by ApJL
The Stokes system with constant viscosity can be cast into different
formulations by exploiting the incompressibility constraint. For instance the
strain in the weak formulation can be replaced by the gradient to decouple the
velocity components in the different coordinate directions. Thus the
discretization of the simplified problem leads to fewer nonzero entries in the
stiffness matrix. This is of particular interest in large scale simulations
where a reduced memory bandwidth requirement can help to significantly
accelerate the computations. In the case of a piecewise constant viscosity, as
it typically arises in multi-phase flows, or when the boundary conditions
involve traction, the situation is more complex, and one has to treat the cross
derivatives in the original Stokes system with care. A naive application of the
standard vectorial Laplacian results in a physically incorrect solution, while
formulations based on the strain increase the computational effort everywhere,
even when the inconsistencies arise only from an incorrect treatment in a small
fraction of the computational domain. Here we propose a new approach that is
consistent with the strain-based formulation and preserves the decoupling
advantages of the gradient-based formulation in isoviscous subdomains. The
modification is equivalent to locally changing the discretization stencils,
hence the more expensive discretization is restricted to a lower dimensional
interface, making the additional computational cost asymptotically negligible.
We demonstrate the consistency and convergence properties of the method and
show that in a massively parallel setup, the multigrid solution of the
resulting discrete systems is faster than for the classical strain-based
formulation. Moreover, we give an application example which is inspired by
geophysical research.
The astronomy community has at its disposal a large back catalog of public
spectroscopic galaxy redshift surveys that can be used for the measurement of
luminosity functions. Utilizing the back catalog with new photometric surveys
to maximum efficiency requires modeling the color selection bias imposed on
selection of target galaxies by flux limits at multiple wavelengths. The
likelihood derived herein can address, in principle, all possible color
selection biases through the use of a generalization of the luminosity
function, $\Phi(L)$, over the space of all spectra: the spectro-luminosity
functional, $\Psi[L_\nu]$. It is, therefore, the first estimator capable of
simultaneously analyzing multiple redshift surveys in a consistent way. We also
propose a new way of parametrizing the evolution of the classic Shechter
function parameters, $L_\star$ and $\phi_\star$, that improves both the
physical realism and statistical performance of the model. The techniques
derived in this work will be used in an upcoming paper to measure the
luminosity function of galaxies at the rest frame wavelength of
$2.4\operatorname{\mu m}$ using the Widefield Infrared Survey Explorer (WISE).
Authors' comments: Accepted to AJ 14 pages, 6 figures
We present new results from a continuing 5 GHz search for flaring radio
emission from a sample of L and T brown dwarfs, conducted with the 305-m
Arecibo radio telescope. In addition to the previously reported flaring from
the T6.5-dwarf 2MASS J10475385+212423, we have detected and confirmed
circularly polarized flares from another T6-dwarf, WISEPC J112254.73+255021.5.
Although the flares are sporadic, they appear to occur at a stable period of
0.288 hours. Given the current constraints, periods equal to its second and
third subharmonic cannot be ruled out. The stability of this period over the
8-month timespan of observations indicates that, if real, it likely reflects
the star's rapid rotation. If confirmed, any of the three inferred
periodicities would be much shorter than the shortest, 1.41-hour rotation
period of a brown dwarf measured so far. This finding would place a new
observational constraint on the angular momentum evolution and rotational
stability of substellar objects. The detection of radio emission from the sixth
~1000 K dwarf further demonstrates that the coolest brown dwarfs and, possibly,
young giant planets, can be efficiently investigated using radio observations
at centimeter wavelengths as a tool.
Authors' comments: Accepted for publication in ApJL; 8 pages, 3 figures
We utilize a Bayesian approach to fit the observed mid-IR-to-submm/mm
spectral energy distributions (SEDs) of 22 WISE-selected and submm-detected,
hyperluminous hot dust-obscured galaxies. By adopting the Torus+GB model, we
decompose the observed IR SEDs of Hot DOGs into torus and cold dust components.
The main results are: 1) Hot DOGs in our submm-detected sample are
hyperluminous, with torus emission dominating the IR energy output. However,
cold dust emission is non-negligible, averagely contributing ~24% of total IR
luminosity. 2) Compared to QSO and starburst SED templates, the median SED of
Hot DOGs shows the highest luminosity ratio between mid-IR and submm at
rest-frame, while it is very similar to that of QSOs at 10-50um suggesting that
the heating sources of Hot DOGs should be buried AGNs. 3) Hot DOGs have both
high dust temperatures ~73K and IR luminosity of cold dust. The T-L relation of
Hot DOGs suggests that the increase in IR luminosity for Hot DOGs is mostly due
to the increase of the dust temperature, rather than dust mass. Hot DOGs have
lower dust masses than those of submillimeter galaxies (SMGs) and QSOs within
the similar redshift range. Both high IR luminosity of cold dust and relatively
low dust mass in Hot DOGs can be expected by their relatively high dust
temperatures. 4) Hot DOGs have high dust covering factors, which deviate the
previously proposed trend of the dust covering factor decreasing with
increasing bolometric luminosity. Finally, we can reproduce the observed
properties in Hot DOGs by employing a physical model of galaxy evolution. The
result suggests that Hot DOGs may lie at or close to peaks of both star
formation and black hole growth histories, and represent a transit phase during
the evolution of massive galaxies, transforming from the dusty starburst
dominated phase to the optically bright QSO phase. (abridged)
Authors' comments: 16 pages, 12 figures. Accepted by ApJ
Layer-wise relevance propagation is a framework which allows to decompose the
prediction of a deep neural network computed over a sample, e.g. an image, down
to relevance scores for the single input dimensions of the sample such as
subpixels of an image. While this approach can be applied directly to
generalized linear mappings, product type non-linearities are not covered. This
paper proposes an approach to extend layer-wise relevance propagation to neural
networks with local renormalization layers, which is a very common product-type
non-linearity in convolutional neural networks. We evaluate the proposed method
for local renormalization layers on the CIFAR-10, Imagenet and MIT Places
datasets.
A majority of early-type galaxies contain interstellar dust, yet the origin
of this dust, and why the dust sometimes exhibits unusual polycyclic aromatic
hydrocarbon (PAH) ratios, remains a mystery. If the dust is internally
produced, it likely originates from the large number of AGB stars associated
with the old stellar population. We present GALEX and WISE elliptical aperture
photometry of $\sim310$ early-type galaxies with Spitzer mid-infrared
spectroscopy and/or ancillary data from ATLAS3D, to characterize their
circumstellar dust and the shape of the radiation field that illuminates the
interstellar PAHs. We find that circumstellar dust is ubiquitous in early-type
galaxies, which indicates some tension between stellar population age estimates
and models for circumstellar dust production in very old stellar populations.
We also use dynamical masses from ATLAS3D to show that WISE W1 (3.4 $\mu$m)
mass-to-light ratios are consistent with the initial mass function variation
found by previous work. While the stellar population differences in early-type
galaxies correspond to a range of radiation field shapes incident upon the
diffuse dust, the ratio of the ionization-sensitive $7.7\mu$m to $11.3\mu$m PAH
feature does not correlate with the shape of the radiation field, nor to
variations with the size-sensitive $11.3\mu$m to $17\mu$m ratio. The $7.7\mu$m
to $11.3\mu$m PAH ratio does tend to be smaller in galaxies with proportionally
greater $H_2$ emission, which is evidence that processing of primarily smaller
grains by shocks is responsible for the unusual ratios, rather than substantial
differences in the overall PAH size or ionization distribution.
Authors' comments: 19 pages, 17 figures, 2 tables, Published in MNRAS
We present WISEA J114724.10$-$204021.3, a young, low-mass, high probability
member of the TW Hya association. WISEA J114724.10$-$204021.3 was discovered
based on its red AllWISE color (W1$-$W2 = 0.63 mag) and extremely red 2MASS
$J-K_{\rm S}$ color ($>$ 2.64 mag), the latter of which is confirmed with
near-infrared photometry from the VISTA Hemisphere Survey ($J-K_{\rm S}$ =
2.57$\pm$0.03). Follow-up near-infrared spectroscopy shows a spectral type of
L7 $\pm$ 1 as well as several spectroscopic indicators of youth. These include
a peaked $H$-band shape and a steeper $K$-band slope, traits typically
attributed to low surface gravity. The sky position, proper motion, and
distance estimates of WISEA J114724.10$-$204021.3 are all consistent with
membership in the $\sim$10 Myr old TW Hya association. Using the age of the TW
Hya association and evolutionary models, we estimate the mass of WISEA
J114724.10$-$204021.3 to be 5$-$13 $M_{\rm Jup}$, making it one of the youngest
and lowest mass free-floating objects yet discovered in the Solar neighborhood.
Authors' comments: Accepted for publication in ApJ Letters
Metal-to-insulator transitions (MITs) are a dramatic manifestation of strong
electron correlations in solids1. The insulating phase can often be suppressed
by quantum tuning, i.e. varying a nonthermal parameter such as chemical
composi- tion or pressure, resulting in a zero-temperature quantum phase
transition (QPT) to a metallic state driven by quantum fluctuations, in
contrast to conventional phase transitions driven by thermal fluctuations.
Theories of exotic phenomena known to occur near the Mott QPT such as quantum
criticality and high-temperature superconductivity often assume a second-order
QPT, but direct experimental evidence for either first- or second-order
behavior at the magnetic QPT associated with the Mott transition has been
scarce and further masked by the superconducting phase in unconventional
superconductors. Most measurements of QPTs have been performed by
volume-integrated probes, such as neutron scattering, magnetization, and
transport, in which discontinuous behavior, phase separation, and spatially
inhomogeneous responses are averaged and smeared out, leading at times to
misidentification as continuous second-order transitions. Here, we demonstrate
through muon spin relaxation/rotation (MuSR) experiments on two archetypal Mott
insulating systems, composition-tuned RENiO3 (RE=rare earth element) and
pressured-tuned V2O3, that the QPT from antiferromagnetic insulator to
paramagnetic metal is first-order: the magnetically ordered volume fraction
decreases to zero at the QPT, resulting in a broad region of intrinsic phase
separation, while the ordered magnetic moment retains its full value across the
phase diagram until it is suddenly destroyed at the QPT. These findings call
for further investigation into the role of inelastic soft modes and the nature
of dynamic spin and charge fluctuations underlying the transition.
Authors' comments: Main text: 9 pages, 3 figures; Supplementary Information: 2 pages, 2
figures
The Near Earth Object Wide-field Infrared Survey Explorer (NEOWISE)
Reactivation mission released data from its first full year of observations in
2015. This data set includes ~2.5 million exposures in each of W1 and W2,
effectively doubling the amount of WISE imaging available at 3.4 and 4.6
microns relative to the AllWISE release. We have created the first ever
full-sky set of coadds combining all publicly available W1 and W2 exposures
from both the AllWISE and NEOWISE-Reactivation (NEOWISER) mission phases. We
employ an adaptation of the unWISE image coaddition framework (Lang 2014),
which preserves the native WISE angular resolution and is optimized for forced
photometry. By incorporating two additional scans of the entire sky, we not
only improve the W1/W2 depths, but also largely eliminate time-dependent
artifacts such as off-axis scattered moonlight. We anticipate that our new
coadds will have a broad range of applications, including target selection for
upcoming spectroscopic cosmology surveys, identification of distant/massive
galaxy clusters, and discovery of high-redshift quasars. In particular, our
full-depth AllWISE+NEOWISER coadds will be an important input for the Dark
Energy Spectroscopic Instrument (DESI) selection of luminous red galaxy and
quasar targets. Our full-depth W1/W2 coadds are already in use within the DECam
Legacy Survey (DECaLS) and Mayall z-band Legacy Survey (MzLS) reduction
pipelines. Much more work still remains in order to fully leverage NEOWISER
imaging for astrophysical applications beyond the solar system.
Authors' comments: coadds available at http://unwise.me, zoomable full-sky rendering at
http://legacysurvey.org/viewer
We propose a novel approach for estimating the location of block boundaries
(change-points) in a random matrix consisting of a block wise constant matrix
observed in white noise. Our method consists in rephrasing this task as a
variable selection issue. We use a penalized least-squares criterion with an
$\ell_1$-type penalty for dealing with this issue. We first provide some
theoretical results ensuring the consistency of our change-point estimators.
Then, we explain how to implement our method in a very efficient way. Finally,
we provide some empirical evidence to support our claims and apply our approach
to HiC data which are used in molecular biology for better understanding the
influence of the chromosomal conformation on the cells functioning.
Authors' comments: 35 pages, 19 figures, submitted
Deep neural networks are capable of modelling highly non-linear functions by
capturing different levels of abstraction of data hierarchically. While
training deep networks, first the system is initialized near a good optimum by
greedy layer-wise unsupervised pre-training. However, with burgeoning data and
increasing dimensions of the architecture, the time complexity of this approach
becomes enormous. Also, greedy pre-training of the layers often turns
detrimental by over-training a layer causing it to lose harmony with the rest
of the network. In this paper a synchronized parallel algorithm for
pre-training deep networks on multi-core machines has been proposed. Different
layers are trained by parallel threads running on different cores with regular
synchronization. Thus the pre-training process becomes faster and chances of
over-training are reduced. This is experimentally validated using a stacked
autoencoder for dimensionality reduction of MNIST handwritten digit database.
The proposed algorithm achieved 26\% speed-up compared to greedy layer-wise
pre-training for achieving the same reconstruction accuracy substantiating its
potential as an alternative.
Compact groups provide an environment to study the growth of galaxies amid
multiple prolonged interactions. With their dense galaxy concentrations and
relatively low velocity dispersions, compact groups mimic the conditions of
hierarchical galaxy assembly. Compact group galaxies are known to show a
bimodality in $Spitzer$ IRAC infrared colorspace: galaxies are preferentially
either quiescent with low specific star formation rates, or are prolifically
forming stars---galaxies with moderate levels of specific star formation are
rare. Previous $Spitzer$ IRAC studies identifying this "canyon" have been
limited by small number statistics. We utilize whole-sky WISE data to study 163
compact groups, thereby tripling our previous sample and including more
galaxies with intermediate mid-IR colors indicative of moderate specific star
formation rates (SSFRs). We define a distinct WISE mid-IR color-space
($\log[{\frac{\rm f_{12}}{\rm f_{4.6}}}]$ vs. $\log[{\frac{\rm f_{22}}{\rm
f_{3.4}}}]$) that we use to identify canyon galaxies from the larger sample. We
confirm that compact group galaxies show a bimodal distribution in the
mid-infrared and identify 37 canyon galaxies with reliable photometry and
intermediate mid-IR colors. Morphologically, we find that the canyon harbors a
large population of both Sa-Sbc and E/S0 type galaxies, and that they fall on
the optical red sequence rather than the green valley. Finally, we provide a
catalog of WISE photometry for 567 of 652 galaxies selected from the sample of
163 compact groups.
Authors' comments: Accepted for publication in The Astrophysical Journal
The Milky Way bulge has a boxy/peanut morphology and an X-shaped structure.
This X-shape has been revealed by the `split in the red clump' from star counts
along the line of sight toward the bulge, measured from photometric surveys.
This boxy, X-shaped bulge morphology is not unique to the Milky Way and such
bulges are observed in other barred spiral galaxies. N-body simulations show
that boxy and X-shaped bulges are formed from the disk via dynamical
instabilities. It has also been proposed that the Milky Way bulge is not
X-shaped, but rather, the apparent split in the red clump stars is a
consequence of different stellar populations, in an old classical spheroidal
bulge. We present a WISE image of the Milky Way bulge, produced by downsampling
the publicly available "unWISE" coadds. The WISE image of the Milky Way bulge
shows that the X-shaped nature of the Milky Way bulge is self-evident and
irrefutable. The X-shape morphology of the bulge in itself and the fraction of
bulge stars that comprise orbits within this structure has important
implications for the formation history of the Milky Way, and, given the
ubiquity of boxy X-shaped bulges, spiral galaxies in general.
We aim to search for hyperliminous infrared (IR) galaxies (HyLIRGs) with IR
luminosity $L_{{\rm IR}}$ $>$ 10$^{13}$ $L_{\odot}$ by applying the selection
method of Dust Obscured Galaxies (DOGs). They are spatially rare but could
correspond to a maximum phase of cosmic star formation and/or active galactic
nucleus (AGN) activity, hence they are a crucial population for understanding
the star formation and mass assembly history of galaxies. Combining the optical
and IR catalogs obtained from Sloan Digital Sky Survey (SDSS) and Wide-field
Infrared Survey Explorer (WISE), we performed the extensive HyLIRGs survey; we
selected 5,311 IR-bright DOGs with $i$ -- [22] $>$ 7.0 and flux at 22 $\mu$m
$>$ 3.8 mJy in 14,555 deg$^2$, where $i$ and [22] are $i$-band and 22 $\mu$m AB
magnitudes, respectively. Among them, 67 DOGs have reliable spectroscopic
redshifts that enable us to estimate their total IR luminosity based on the SED
fitting. Consequently, we successfully discovered 24 HyLIRGs among the 67
spectroscopically-confirmed DOGs. We found that (i) $i$ - [22] color of
IR-bright DOGs correlates with the total IR luminosity and (ii) surface number
density of HyLIRGs is $>$ 0.17 deg$^{-2}$. A high fraction ($\sim$ 73%) of
IR-bright DOGs with $i$ - [22] $>$ 7.5 shows $L_{{\rm IR}}$ $>$ 10$^{13}$
$L_{\odot}$, and the DOGs criterion we adopted could be independently-effective
against "W1W2-dropout method" based on four WISE bands, for searching hyper IR
luminous populations of galaxies.
Authors' comments: 10 pages and 14 figures, accepted for publication in ApJ
2637.
The First Detection of Photometric Variability in a Y Dwarf: WISE
J140518.39+553421.3
Michael C. Cushing, Kevin K. Hardegree-Ullman, Jesica L. Trucks, Caroline V. Morley, John E. Gizis, Mark S. Marley, Jonathan J. Fortney, J. Davy Kirkpatrick et al.
We present the first detection of photometric variability of a
spectroscopically-confirmed Y dwarf. The Infrared Array Camera on board the
Spitzer Space Telescope was used to obtain times series photometry at 3.6 and
4.5 microns over a twenty four hour period at two different epochs separated by
149 days. Variability is evident at 4.5 um in the first epoch and at 3.6 and
4.5 um in the second epoch which suggests that the underlying cause or causes
of this variability change on the timescales of months. The second-epoch [3.6]
and [4.5] light curves are nearly sinusoidal in form, in phase, have periods of
roughly 8.5 hours, and have semi-amplitudes of 3.5%. We find that a simple
geometric spot model with a single bright spot reproduces these observations
well. We also compare our measured semi-amplitudes of the second epoch light
curves to predictions of the static, one-dimensional, partly cloudy and hot
spot models of Morley and collaborators and find that neither set of models can
reproduce the observed [3.6] and[4.5] semi-amplitudes simultaneously. More
advanced two- or three-dimensional models that include time-dependent phenomena
like vertical mixing, cloud formation, and thermal relaxation are therefore
sorely needed in order to properly interpret our observations.
Authors' comments: Accepted for publication in the Astrophysical Journal
We explored the AllWISE catalogue of the Wide-field Infrared Survey Explorer
mission and identified Young Stellar Object candidates. Reliable 2MASS and WISE
photometric data combined with Planck dust opacity values were used to build
our dataset and to find the best classification scheme. A sophisticated
statistical method, the Support Vector Machine (SVM) is used to analyse the
multi-dimensional data space and to remove source types identified as
contaminants (extragalactic sources, main sequence stars, evolved stars and
sources related to the interstellar medium). Objects listed in the SIMBAD
database are used to identify the already known sources and to train our
method. A new all-sky selection of 133,980 Class I/II YSO candidates is
presented. The estimated contamination was found to be well below 1% based on
comparison with our SIMBAD training set. We also compare our results to that of
existing methods and catalogues. The SVM selection process successfully
identified >90% of the Class I/II YSOs based on comparison with photometric and
spectroscopic YSO catalogues. Our conclusion is that by using the SVM, our
classification is able to identify more known YSOs of the training sample than
other methods based on colour-colour and magnitude-colour selection. The
distribution of the YSO candidates well correlates with that of the Planck
Galactic Cold Clumps in the Taurus--Auriga--Perseus--California region.
Authors' comments: 27 pages, 15 figures, 15 tables
We report on experimentally observed addition, subtraction and cancellation
of orbital angular momentum (OAM) in the process of parametric four-wave mixing
that results in frequency up- and down-converted emission in Rb vapour.
Specific features of OAM transfer from resonant laser fields with different
optical topological charges to the spatially and temporally coherent blue light
have been considered. We have observed the conservation of OAM in nonlinear
wave mixing in a wide range of experimental conditions including a
non-collinear geometry of the applied laser beams and furthermore that the
coherent blue light accumulates the total orbital angular momentum of the
applied laser light. Spectral and power dependences of vortex and plane wave
front blue light beams have been compared.
Authors' comments: 4 pages, 6 figures
Throughout the last decade sensitive infrared observations obtained by the
Spitzer Space Telescope significantly increased the known population of YSOs
associated with nearby molecular clouds. With such a census recent studies have
characterized pre-main sequence stars (PMS) and determined parame- ters from
different wavelengths. Given the restricted Spitzer coverage of some of these
clouds, relative to their extended regions, these YSO populations may represent
a limited view of star formation in these regions. We are taking advantage of
mid-infrared observations from the NASA Wide Field Infrared Survey Explorer
(WISE), which provides an all sky view and therefore full coverage of the
nearby clouds, to assess the degree to which their currently known YSO
population may be representative of a more complete population. We extend the
well established classification method of the Spitzer Legacy teams to archived
WISE observations. We have adopted 2MASS photometry as a standard catalogue for
comparisons. Besides the massive embedded cluster IC5146 we provide a multiband
view of five new embedded clusters in its surroundings that we discovered with
WISE. In short, the analysis in- volves the following for the presently studied
cluster sample: (1) extraction of 2MASS/WISE/Spitzer photometry in a wide
circular region; (2) field-star decontamination to enhance the intrinsic
Colour-magnitude diagram (CMD) morphology (essential for a proper derivation of
reddening, age, and distance from the Sun); and (3) construction of
Colour-magnitude filters, for more contrasted stellar radial density profiles
(RDPs).
Authors' comments: arXiv admin note: text overlap with arXiv:astro-ph/0306595,
arXiv:astro-ph/0406003 by other authors
