In the expanding universe, relativistic scalar fields are thought to be attenuated by Hubble friction, which results from the dilation of the underlying spacetime metric. By contrast, in a contracting universe this pseudo-friction would lead to ampli
fication. Here, we experimentally measure both Hubble attenuation and amplification in expanding and contracting toroidally-shaped Bose-Einstein condensates, in which phonons are analogous to cosmological scalar fields. We find that the observed attenuation or amplification depends on the temporal phase of the phonon field, which is only possible for non-adiabatic dynamics, in contrast to the expanding universe in its current epoch, which is adiabatic. The measured strength of the Hubble friction disagrees with recent theory [J. M. Gomez Llorente and J. Plata, Phys. Rev. A 100 043613 (2019) and S. Eckel and T. Jacobson, SciPost Phys. 10 64 (2021)], suggesting that our model does not yet capture all relevant physics. While our current work focuses on coherent-state phonons, it can be extended to regimes where quantum fluctuations in causally disconnected regions of space become important, leading to spontaneous pair-production.
In this work, we studied the impact of galaxy morphology on photometric redshift (photo-$z$) probability density functions (PDFs). By including galaxy morphological parameters like the radius, axis-ratio, surface brightness and the Sersic index in ad
dition to the $ugriz$ broadbands as input parameters, we used the machine learning photo-$z$ algorithm ANNz2 to train and test on galaxies from the Canada-France-Hawaii Telescope Stripe-82 (CS82) Survey. Metrics like the continuous ranked probability score (CRPS), probability integral transform (PIT), Bayesian odds parameter, and even the width and height of the PDFs were evaluated, and the results were compared when different number of input parameters were used during the training process. We find improvements in the CRPS and width of the PDFs when galaxy morphology has been added to the training, and the improvement is larger especially when the number of broadband magnitudes are lacking.
Chromospheric umbral oscillations produce periodic brightenings in the core of some spectral lines, known as umbral flashes. They are also accompanied by fluctuations in velocity, temperature, and, according to several recent works, magnetic field. I
n this study, we aim to ascertain the accuracy of the magnetic field determined from
Resource management plays a pivotal role in wireless networks, which, unfortunately, leads to challenging NP-hard problems. Artificial Intelligence (AI), especially deep learning techniques, has recently emerged as a disruptive technology to solve su
ch challenging problems in a real-time manner. However, although promising results have been reported, practical design guidelines and performance guarantees of AI-based approaches are still missing. In this paper, we endeavor to address two fundamental questions: 1) What are the main advantages of AI-based methods compared with classical techniques; and 2) Which neural network should we choose for a given resource management task. For the first question, four advantages are identified and discussed. For the second question, emph{optimality gap}, i.e., the gap to the optimal performance, is proposed as a measure for selecting model architectures, as well as, for enabling a theoretical comparison between different AI-based approaches. Specifically, for $K$-user interference management problem, we theoretically show that graph neural networks (GNNs) are superior to multi-layer perceptrons (MLPs), and the performance gap between these two methods grows with $sqrt{K}$.
Various attempts have been made in the literature at describing the origin and the physical mechanisms behind flaring events in blazars with radiative emission models, but detailed properties of multi-wavelength (MWL) light curves still remain diffic
ult to reproduce. We have developed a versatile radiative code, based on a time-dependent treatment of particle acceleration, escape and radiative cooling, allowing us to test different scenarios to connect the continuous low-state emission self-consistently with that during flaring states. We consider flares as weak perturbations of the quiescent state and apply this description to the February 2010 MWL flare of Mrk 421, the brightest Very High Energy (VHE) flare ever detected from this archetypal blazar, focusing on interpretations with a minimum number of free parameters. A general criterion is obtained, which disfavours a one-zone model connecting low and high state under our assumptions. A two-zone model combining physically connected acceleration and emission regions yields a satisfactory interpretation of the available time-dependent MWL light curves and spectra of Mrk 421, although certain details remain difficult to reproduce. The two-zone scenario finally proposed for the complex quiescent and flaring VHE emitting region involves both Fermi-I and Fermi-II acceleration mechanisms, respectively at the origin of the quiescent and flaring emission.
Sparse Bayesian learning (SBL) is a powerful framework for tackling the sparse coding problem while also providing uncertainty quantification. However, the most popular inference algorithms for SBL become too expensive for high-dimensional problems d
ue to the need to maintain a large covariance matrix. To resolve this issue, we introduce a new SBL inference algorithm that avoids explicit computation of the covariance matrix, thereby saving significant time and space. Instead of performing costly matrix
Neural networks (NNs) are often used as surrogates or emulators of partial differential equations (PDEs) that describe the dynamics of complex systems. A virtually negligible computational cost of such surrogates renders them an attractive tool for e
nsemble-based computation, which requires a large number of repeated PDE solves. Since the latter are also needed to generate sufficient data for NN training, the usefulness of NN-based surrogates hinges on the balance between the training cost and the computational gain stemming from their deployment. We rely on multi-fidelity simulations to reduce the cost of data generation for subsequent training of a deep convolutional NN (CNN) using transfer learning. High- and low-fidelity images are generated by solving PDEs on fine and coarse meshes, respectively. We use theoretical results for multilevel Monte Carlo to guide our choice of the numbers of images of each kind. We demonstrate the performance of this multi-fidelity training strategy on the problem of estimation of the distribution of a quantity of interest, whose dynamics is governed by a system of nonlinear PDEs (parabolic PDEs of multi-phase flow in heterogeneous porous media) with uncertain/random parameters. Our numerical experiments demonstrate that a mixture of a comparatively large number of low-fidelity data and smaller numbers of high- and low-fidelity data provides an optimal balance of computational speed-up and prediction accuracy. The former is reported relative to both CNN training on high-fidelity images only and Monte Carlo solution of the PDEs. The latter is expressed in terms of both the Wasserstein distance and the Kullback-Leibler divergence.
Intelligent reflecting surface (IRS) is a promising enabler for next-generation wireless communications due to its reconfigurability and high energy efficiency in improving the propagation condition of channels. In this paper, we consider a large-sca
le IRS-aided multiple-input-multiple-output (MIMO) communication system in which statistical channel state information (CSI) is available at the transmitter. By leveraging random matrix theory, we first derive a deterministic approximation (DA) of the ergodic rate with low computation complexity and prove the existence and uniqueness of the DA parameters. Then, we propose an alternating optimization algorithm to obtain a locally optimal solution for maximizing the DA with respect to phase shifts and signal covariance matrices. Numerical results will show that the DA is tight and our proposed method can improve the ergodic rate effectively.
Convolutional dictionary learning (CDL), the problem of estimating shift-invariant templates from data, is typically conducted in the absence of a prior/structure on the templates. In data-scarce or low signal-to-noise ratio (SNR) regimes, which have
received little attention from the community, learned templates overfit the data and lack smoothness, which can affect the predictive performance of downstream tasks. To address this limitation, we propose GPCDL, a convolutional dictionary learning framework that enforces priors on templates using Gaussian Processes (GPs). With the focus on smoothness, we show theoretically that imposing a GP prior is equivalent to Wiener filtering the learned templates, thereby suppressing high-frequency components and promoting smoothness. We show that the algorithm is a simple extension of the classical iteratively reweighted least squares, which allows the flexibility to experiment with different smoothness assumptions. Through simulation, we show that GPCDL learns smooth dictionaries with better accuracy than the unregularized alternative across a range of SNRs. Through an application to neural spiking data from rats, we show that learning templates by GPCDL results in a more accurate and visually-interpretable smooth dictionary, leading to superior predictive performance compared to non-regularized CDL, as well as parametric alternatives.
We investigate the time-dependent perturbations of strongly coupled $mathcal{N} = 4$ SYM theory at finite temperature and finite chemical potential with a second order phase transition. This theory is modelled by a top-down Einstein-Maxwell-dilaton d
escription which is a consistent truncation of the dimensional reduction of type IIB string theory on AdS$_5times$S$^5$. We focus on spin-1 and spin-2 sectors of perturbations and compute the linearized hydrodynamic transport coefficients up to the third order in gradient expansion. We also determine the radius of convergence of the hydrodynamic mode in spin-1 sector and the lowest non-hydrodynamic modes in spin-2 sector. Analytically, we find that all the hydrodynamic quantities have the same critical exponent near the critical point $theta = 1/2$. Moreover, we establish a relation between symmetry enhancement of the underlying theory and vanishing the only third order hydrodynamic transport coefficient $theta_1$, which appears in the shear dispersion relation of a conformal theory on a flat background.
