Search

Researchers have proposed many methods for fair and robust machine learning, but comprehensive empirical evaluation of their subgroup robustness is lacking. In this work, we address this gap in the context of tabular data, where sensitive subgroups are clearly-defined, real-world fairness problems abound, and prior works often do not compare to state-of-the-art tree-based models as baselines. We conduct an empirical comparison of several previously-proposed methods for fair and robust learning alongside state-of-the-art tree-based methods and other baselines. Via experiments with more than $340{,}000$ model configurations on eight datasets, we show that tree-based methods have strong subgroup robustness, even when compared to robustness- and fairness-enhancing methods. Moreover, the best tree-based models tend to show good performance over a range of metrics, while robust or group-fair models can show brittleness, with significant performance differences across different metrics for a fixed model. We also demonstrate that tree-based models show less sensitivity to hyperparameter configurations, and are less costly to train. Our work suggests that tree-based ensemble models make an effective baseline for tabular data, and are a sensible default when subgroup robustness is desired. For associated code and detailed results, see https://github.com/jpgard/subgroup-robustness-grows-on-trees ., Comment: To appear at Neural Information Processing Systems (NeurIPS) 2022. Code at https://github.com/jpgard/subgroup-robustness-grows-on-trees

Extraction of information in the form of oscillations from noisy data of natural phenomena such as sounds, earthquakes, ionospheric and brain activity, and various emissions from cosmic objects is extremely difficult. As a method for finding periodicity in such challenging data sets, the 2D Hybrid approach, which employs wavelets, is presented. Our technique produces a wavelet transform correlation intensity contour map for two (or one) time series on a period plane defined by two independent period axes. Notably, by spreading peaks across the second dimension, our method improves apparent resolution of detected oscillations in the period plane and identifies the direction of signal changes using correlation coefficients. We demonstrate the performance of the 2D Hybrid technique on a very low frequency (VLF) signal emitted in Italy and recorded in Serbia in time vicinity of the occurrence of an earthquake on November 3, 2010, near Kraljevo, Serbia. We identified a distinct signal in the range 120-130 s that appears only in association with the considered earthquake. Other wavelets, such as Superlets, which may detect fast transient oscillations, will be employed in the future analysis., Comment: published in Mathematics MDPI

Fluid pumping and the generation of electric current by living tissues are required during morphogenetic processes and for maintainance of homeostasis. How these flows emerge from active and passive ion transport in cells has been well established. However, the interplay between flow and current generation is not well understood. Here, we study the electro-hydraulic coupling that arises from cell ion pumping. We develop a one-dimensional continuum model of fluid and ion transport across active cell membranes. Solving the Nernst-Planck and Poisson equations in the limit of weak charge imbalance allows us to derive approximate analytical solutions of the model. These approximations, consistent with the numerical results in physiologically relevant regime of parameters, allow us to describe electro-hydraulic activity of cells and tissues in terms of experimentally accessible parameters.

Classification of cancer cellularity within tissue samples is currently a manual process performed by pathologists. This process of correctly determining cancer cellularity can be time intensive. Deep Learning (DL) techniques in particular have become increasingly more popular for this purpose, due to the accuracy and performance they exhibit, which can be comparable to the pathologists. This work investigates the capabilities of two DL approaches to assess cancer cellularity in whole slide images (WSI) in the SPIE-AAPM-NCI BreastPathQ challenge dataset. The effects of training on augmented data via rotations, and combinations of multiple architectures into a single network were analyzed using a modified Kendall Tau-b prediction probability metric known as the average prediction probability PK. A deep, transfer learned, Convolutional Neural Network (CNN) InceptionV3 was used as a baseline, achieving an average PK value of 0.884, showing improvement from the average PK value of 0.83 achieved by pathologists. The network was then trained on additional training datasets which were rotated between 1 and 360 degrees, which saw a peak increase of PK up to 4.2%. An additional architecture consisting of the InceptionV3 network and VGG16, a shallow, transfer learned CNN, was combined in a parallel architecture. This parallel architecture achieved a baseline average PK value of 0.907, a statistically significantly improvement over either of the architectures' performances separately (p<0.0001 by unpaired t-test)., Comment: 7 pages (includes a cover page), 5 figures, 1 table, work addresses the BreastPathQ challenge

Deep neural network (DNN)-assisted channel coding designs, such as low-complexity neural decoders for existing codes, or end-to-end neural-network-based auto-encoder designs are gaining interest recently due to their improved performance and flexibility; particularly for communication scenarios in which high-performing structured code designs do not exist. Communication in the presence of feedback is one such communication scenario, and practical code design for feedback channels has remained an open challenge in coding theory for many decades. Recently, DNN-based designs have shown impressive results in exploiting feedback. In particular, generalized block attention feedback (GBAF) codes, which utilizes the popular transformer architecture, achieved significant improvement in terms of the block error rate (BLER) performance. However, previous works have focused mainly on passive feedback, where the transmitter observes a noisy version of the signal at the receiver. In this work, we show that GBAF codes can also be used for channels with active feedback. We implement a pair of transformer architectures, at the transmitter and the receiver, which interact with each other sequentially, and achieve a new state-of-the-art BLER performance, especially in the low SNR regime.

In this paper, the possibility of determining the thermal, elastic and geometric characteristics of a thin TiO2 film deposited on a silicon substrate, thickness 30 mikrons, in the frequency range of 20 to 20 kHz with neural networks was analyzed. For this purpose, the substrate parameters remained the known and constant in the two-layer model and nano layer thin-film parameters were changed: thickness, expansion and thermal diffusivity. Prediction of these three parameters was analyzed separately with three neural networks and all of these together by fourth neural network. It was shown that neural network, which analyzed all three parameters at the same time, achieved the highest accuracy, so the use of networks that provide predictions for only one parameter is less reliable., Comment: 21 pages, 5 figures

We investigate how randomly oriented cell traction forces lead to fluidisation in a vertex model of epithelial tissues. We find that the fluidisation occurs at a critical value of the traction force magnitude $F_c$. We show that this transition exhibits critical behaviour, similar to the yielding transition of sheared amorphous solids. However, we find that it belongs to a different universality class, even though it satisfies the same scaling relations between critical exponents established in the yielding transition of sheared amorphous solids. Our work provides a fluidisation mechanism through active force generation that could be relevant in biological tissues.

The $Euclid$ mission will provide first-of-its-kind coverage in the near-infrared over deep (three fields, $\sim$10-20 square degrees each) and wide ($\sim$10000 square degrees) fields. While the survey is not designed to discover transients, the deep fields will have repeated observations over a two-week span, followed by a gap of roughly six months. In this analysis, we explore how useful the deep field observations will be for measuring properties of Type Ia supernovae (SNe Ia). Using simulations that include $Euclid$'s planned depth, area and cadence in the deep fields, we calculate that more than 3700 SNe between $0.00.8$. Furthermore, we predict how well additional $Euclid$ mock data can be used to constrain a key systematic in SN Ia studies - the size of the luminosity 'step' found between SNe hosted in high mass ($>10^{10} M_{\odot}$) and low mass ($>10^{10} M_{\odot}$) galaxies. This measurement has unique information in the rest-frame NIR. We predict that if the step is caused by dust, we will be able to measure its reduction in the NIR compared to optical at the 4$\sigma$ level. We highlight that the LSST and $Euclid$ observing strategies used in this work are still provisional and some level of joint processing is required. Still, these first results are promising, and assuming $Euclid$ begins observations well before the Nancy Roman Space Telescope (Roman), we expect this dataset to be extremely helpful for preparation for Roman itself.

Multiple Coulomb Scattering (MCS) is a well known phenomenon occurring when charged particles traverse materials. Measurements of muons traversing low $Z$ materials made in the MuScat experiment showed that theoretical models and simulation codes, such as GEANT4 (v7.0), over-estimated the scattering. The Muon Ionization Cooling Experiment (MICE) measured the cooling of a muon beam traversing a liquid hydrogen or lithium hydride (LiH) energy absorber as part of a programme to develop muon accelerator facilities, such as a Neutrino Factory or a Muon Collider. The energy loss and MCS that occur in the absorber material are competing effects that alter the performance of the cooling channel. Therefore measurements of MCS are required in order to validate the simulations used to predict the cooling performance in future accelerator facilities. We report measurements made in the MICE apparatus of MCS using a LiH absorber and muons within the momentum range 160 to 245 MeV/c. The measured RMS scattering width is about 9% smaller than that predicted by the approximate formula proposed by the Particle Data Group. Data at 172, 200 and 240 MeV/c are compared to the GEANT4 (v9.6) default scattering model. These measurements show agreement with this more recent GEANT4 (v9.6) version over the range of incident muon momenta., Comment: 20 pages, 14 figures, journal

The vibrational properties of ferrimagnetic $\mathrm{Mn_3Si_{2}Te_6}$ single crystals are investigated using Raman spectroscopy and density functional theory calculations. Eighteen Raman-active modes are identified, fourteen of which are assigned according to with the trigonal symmetry. Four additional peaks, obeying the $A_{1g}$ selection rules, are attributed to the overtones. The unconventional temperature evolution of the $A_{1g}^5$ mode self-energy suggests a competition between different short-range magnetic correlations that significantly impact the spin-phonon interaction in $\mathrm{Mn_3Si_{2}Te_6}$. The research provides a comprehensive insight to the lattice properties, studies their temperature dependence and shows the arguments for existence of competing short-range magnetic phases in $\mathrm{Mn_3Si_{2}Te_6}$., Comment: 6 pages, 3 figures, 2 tables

Microwave receivers using electromagnetically-induced transparency (EIT) in Rydberg atoms have recently demonstrated improved sensitivities. It is not evident how their state-of-the-art electric field sensitivities compare to those achieved using standard electronic receivers consisting of low-noise amplifiers (LNAs) and mixers. In this paper, we show that conventional room-temperature electronic receivers greatly outperform the best demonstrated sensitivities of room-temperature Rydberg electrometers in standard free-space coupled configurations. However, Rydberg-atom receivers can surpass the sensitivity of conventional receivers if resonant or confining microwave structures are designed to enhance the electric fields sensed by the atoms. For a given microwave resonator, the external (coupling) quality factor must be carefully chosen to minimize their thermal and quantum noise contributions. Closed-form expressions for these optimal design points are found, and compared in terms of noise temperature with conventional LNAs reported in the literature from 600 MHz to 330 GHz., Comment: Fix typos on Fig. 16. Add refs [14],[16],[17]. Reword abstract

We study delayed loss of stability in a class of fast-slow systems with two fast variables and one slow one, where the linearisation of the fast vector field along a one-dimensional critical manifold has two real eigenvalues which intersect before the accumulated contraction and expansion are balanced along any individual eigendirection. That interplay between eigenvalues and eigendirections renders the use of known entry-exit relations unsuitable for calculating the point at which trajectories exit neighbourhoods of the given manifold. We illustrate the various qualitative scenarios that are possible in the class of systems considered here, and we propose novel formulae for the entry-exit functions that underlie the phenomenon of delayed loss of stability therein.

Regression discontinuity design (RDD) is a quasi-experimental approach to study the causal effects of an intervention/treatment on later health outcomes. It exploits a continuously measured assignment variable with a clearly defined cut-off above or below which the population is at least partially assigned to the intervention/treatment. We describe the RDD and outline the applications of RDD in the context of perinatal epidemiology and birth cohort research. There is an increasing number of studies using RDD in perinatal and pediatric epidemiology. Most of these studies were conducted in the context of education, social and welfare policies, healthcare organization, insurance, and preventive programs. Additional thematic fields include clinically relevant research questions, shock events, social and environmental factors, and changes in guidelines. Maternal and perinatal characteristics, such as age, birth weight and gestational age are frequently used assignment variables to study the effects of the type and intensity of neonatal care, health insurance, and supplemental newborn benefits. Different socioeconomic measures have been used to study the effects of social, welfare and cash transfer programs, while age or date of birth served as assignment variables to study the effects of vaccination programs, pregnancy-specific guidelines, maternity and paternity leave policies and introduction of newborn-based welfare programs. RDD has advantages, including relatively weak and testable assumptions, strong internal validity, intuitive interpretation, and transparent and simple graphical representation. However, its use in birth cohort research is hampered by the rarity of settings outside of policy and program evaluations, low statistical power, limited external validity (geographic- and time-specific settings) and potential contamination by other exposures/interventions., Comment: Main text: 32 pages, 6 figures, 2 tables, Supplementary material: 24 pages

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will detect an unprecedentedly large sample of actively accreting supermassive black holes with typical accretion disk (AD) sizes of a few light days. This brings us to face challenges in the reverberation mapping (RM) measurement of AD sizes in active galactic nuclei (AGNs) using interband continuum delays. We examine the effect of LSST cadence strategies on AD RM using our metric AGNTimeLagMetric. It accounts for redshift, cadence, the magnitude limit, and magnitude corrections for dust extinction. Running our metric on different LSST cadence strategies, we produce an atlas of the performance estimations for LSST photometric RM measurements. We provide an upper limit on the estimated number of quasars for which the AD time lag can be computed within 01000 sources in each Deep Drilling field (DDF, 10 sq. deg) in any filter, with the redshift distribution of these sources peaking at z~1. We find the LSST observation strategies with a good cadence (~ 5 days) and a long cumulative season (~9 yr), as proposed for LSST DDF, are favored for the AD size measurement. We create synthetic LSST light curves for the most suitable DDF cadences and determine RM time lags to demonstrate the impact of the best cadences based on the proposed metric., Comment: accepted for publication in the Astrophysical Journal Focus Issue on Rubin LSST cadence and survey strategy

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) dataset will dramatically alter our understanding of the Universe, from the origins of the Solar System to the nature of dark matter and dark energy. Much of this research will depend on the existence of robust, tested, and scalable algorithms, software, and services. Identifying and developing such tools ahead of time has the potential to significantly accelerate the delivery of early science from LSST. Developing these collaboratively, and making them broadly available, can enable more inclusive and equitable collaboration on LSST science. To facilitate such opportunities, a community workshop entitled "From Data to Software to Science with the Rubin Observatory LSST" was organized by the LSST Interdisciplinary Network for Collaboration and Computing (LINCC) and partners, and held at the Flatiron Institute in New York, March 28-30th 2022. The workshop included over 50 in-person attendees invited from over 300 applications. It identified seven key software areas of need: (i) scalable cross-matching and distributed joining of catalogs, (ii) robust photometric redshift determination, (iii) software for determination of selection functions, (iv) frameworks for scalable time-series analyses, (v) services for image access and reprocessing at scale, (vi) object image access (cutouts) and analysis at scale, and (vii) scalable job execution systems. This white paper summarizes the discussions of this workshop. It considers the motivating science use cases, identified cross-cutting algorithms, software, and services, their high-level technical specifications, and the principles of inclusive collaborations needed to develop them. We provide it as a useful roadmap of needs, as well as to spur action and collaboration between groups and individuals looking to develop reusable software for early LSST science., Comment: White paper from "From Data to Software to Science with the Rubin Observatory LSST" workshop

Recent analyses have found intriguing correlations between the colour ($c$) of type Ia supernovae (SNe Ia) and the size of their 'mass-step', the relationship between SN Ia host galaxy stellar mass ($M_\mathrm{stellar}$) and SN Ia Hubble residual, and suggest that the cause of this relationship is dust. Using 675 photometrically-classified SNe Ia from the Dark Energy Survey 5-year sample, we study the differences in Hubble residual for a variety of global host galaxy and local environmental properties for SN Ia subsamples split by their colour. We find a $3\sigma$ difference in the mass-step when comparing blue ($c<0$) and red ($c>0$) SNe. We observe the lowest r.m.s. scatter ($\sim0.14$ mag) in the Hubble residual for blue SNe in low mass/blue environments, suggesting that this is the most homogeneous sample for cosmological analyses. By fitting for $c$-dependent relationships between Hubble residuals and $M_\mathrm{stellar}$, approximating existing dust models, we remove the mass-step from the data and find tentative $\sim 2\sigma$ residual steps in rest-frame galaxy $U-R$ colour. This indicates that dust modelling based on $M_\mathrm{stellar}$ may not fully explain the remaining dispersion in SN Ia luminosity. Instead, accounting for a $c$-dependent relationship between Hubble residuals and global $U-R$, results in $\leq1\sigma$ residual steps in $M_\mathrm{stellar}$ and local $U-R$, suggesting that $U-R$ provides different information about the environment of SNe Ia compared to $M_\mathrm{stellar}$, and motivating the inclusion of galaxy $U-R$ colour in SN Ia distance bias correction., Comment: 19 pages, 8 figures. Published in MNRAS

Safe motion planning in robotics requires planning into space which has been verified to be free of obstacles. However, obtaining such environment representations using lidars is challenging by virtue of the sparsity of their depth measurements. We present a learning-aided 3D lidar reconstruction framework that upsamples sparse lidar depth measurements with the aid of overlapping camera images so as to generate denser reconstructions with more definitively free space than can be achieved with the raw lidar measurements alone. We use a neural network with an encoder-decoder structure to predict dense depth images along with depth uncertainty estimates which are fused using a volumetric mapping system. We conduct experiments on real-world outdoor datasets captured using a handheld sensing device and a legged robot. Using input data from a 16-beam lidar mapping a building network, our experiments showed that the amount of estimated free space was increased by more than 40% with our approach. We also show that our approach trained on a synthetic dataset generalises well to real-world outdoor scenes without additional fine-tuning. Finally, we demonstrate how motion planning tasks can benefit from these denser reconstructions., Comment: Accepted at IROS 2022; Video at https://www.youtube.com/watch?v=db3HetMq5h4

We present the luminosity functions and host galaxy properties of the Dark Energy Survey (DES) core-collapse supernova (CCSN) sample, consisting of 69 Type II and 50 Type Ibc spectroscopically and photometrically-confirmed supernovae over a redshift range $0.045

We study a singularly perturbed fast-slow system of two partial differential equations (PDEs) of reaction-diffusion type on a bounded domain via Galerkin discretisation. We assume that the reaction kinetics in the fast variable realise a generic fold singularity, whereas the slow variable takes the role of a dynamic bifurcation parameter, thus extending the classical analysis of the singularly perturbed fold. Our approach combines a spectral Galerkin discretisation with techniques from geometric singular perturbation theory which are applied to the resulting high-dimensional systems of ordinary differential equations. In particular, we show the existence of invariant slow manifolds in the phase space of the original system of PDEs away from the fold singularity, while the passage past the singularity of the Galerkin manifolds obtained after discretisation is described by geometric desingularisation, or blow-up. Finally, we discuss the relation between these Galerkin manifolds and the underlying slow manifolds., Comment: 45 pages with 9 figures. Corresponding author: Thomas Zacharis

Type Ia supernovae (SNe Ia) are used as standardisable candles to measure cosmological distances, but differences remain in their corrected luminosities which display a magnitude step as a function of host galaxy properties such as stellar mass and rest-frame $U-R$ colour. Identifying the cause of these steps is key to cosmological analyses and provides insight into SN physics. Here we investigate the effects of SN progenitor ages on their light curve properties using a galaxy-based forward model that we compare to the Dark Energy Survey 5-year SN Ia sample. We trace SN Ia progenitors through time and draw their light-curve width parameters from a bimodal distribution according to their age. We find that an intrinsic luminosity difference between SNe of different ages cannot explain the observed trend between step size and SN colour. The data split by stellar mass are better reproduced by following recent work implementing a step in total-to-selective dust extinction ratio $(R_V)$ between low- and high-mass hosts, although an additional intrinsic luminosity step is still required to explain the data split by host galaxy $U-R$. Modelling the $R_V$ step as a function of galaxy age provides a better match overall. Additional age vs. luminosity steps marginally improve the match to the data, although most of the step is absorbed by the width vs. luminosity coefficient $\alpha$. Furthermore, we find no evidence that $\alpha$ varies with SN age., Comment: 20 pages, 16 figures. Accepted for publication in MNRAS

We study the three-timescale dynamics of a model that describes the El Ni\~no Southern Oscillation (ENSO) phenomenon, which was proposed in [A. Roberts, J. Guckenheimer, E. Widiasih, A. Timmermann, and C. K. Jones, Mixed-mode oscillations of El Ni\~no--Southern Oscillation, Journal of the Atmospheric Sciences, 73 (2016), pp. 1755--1766]. While ENSO phenomena are inherently characterised by the presence of multiple distinct timescales, the above model has previously been studied in a two-timescale context only. Here, we uncover the geometric mechanisms that are responsible for complex oscillatory dynamics in a three-timescale regime, and we demonstrate that the system exhibits a variety of qualitatively different behaviours, such as mixed-mode oscillation (MMO) with ``plateaus" -- trajectories where epochs of quiescence alternate with dramatic excursions -- and relaxation oscillation. The latter, although emergent also in the two-timescale context in appropriate parameter regimes, had not been documented previously. Moreover, we show that these mechanisms are relevant to models from other fields of ecological and population dynamics, as the underlying geometry is similar to the unfolding of Rosenzweig--MacArthur-type models in three dimensions.

Type Ia supernovae (SNe Ia) are powerful tools for measuring the expansion history of the universe, but the impact of dust around SNe Ia remains unknown and is a critical systematic uncertainty. One way to improve our empirical description of dust is to analyse highly reddened SNe Ia ($E(B-V)>0.4$, roughly equivalent to the fitted SALT2 light-curve parameter $c>0.3$). With the recently released Pantheon+ sample, there are 57 SNe Ia that were removed because of their high colour alone (with colours up to $c=1.61$), which can provide enormous leverage on understanding line-of-sight $R_V$. Previous studies have claimed that $R_V$ decreases with redder colour, though it is unclear if this is due to limited statistics, selection effects, or an alternative explanation. To test this claim, we fit two separate colour-luminosity relationships, one for the main cosmological sample ($c<0.3$) and one for highly reddened ($c>0.3$) SNe Ia. We find the change in the colour-luminosity coefficient to be consistent with zero. Additionally, we compare the data to simulations with different colour models, and find that the data prefers a model with a flat dependence of $R_V$ on colour over a declining dependence. Finally, our results strongly support that line-of-sight $R_V$ to SNe Ia is not a single value, but forms a distribution., Comment: Accepted to MNRAS, updating to match accepted version. 11 pages, 9 figures, 5 tables

1. Sample size estimation through power analysis is a fundamental tool in planning an ecological study, yet there are currently no well-established procedures for when multivariate abundances are to be collected. A power analysis procedure would need to address three challenges: designing a parsimonious simulation model that captures key community data properties; measuring effect size in a realistic yet interpretable fashion; and ensuring computational feasibility when simulation is used both for power estimation and significance testing. 2. Here we propose a power analysis procedure that addresses these three challenges by: using for simulation a Gaussian copula model with factor analytical structure, fitted to pilot data; assuming a common effect size across all taxa, but applied in different directions according to expert opinion (to "increaser", "decreaser" or "no effect" taxa); using a critical value approach to estimate power, which reduces computation time by a factor of 500 with little loss of accuracy. 3. The procedure is demonstrated on pilot data from fish assemblages in a restoration study, where it was found that the planned study design would only be capable of detecting relatively large effects (change in abundance by a factor of 1.5 or more). 4. The methods outlined in this paper are available in accompanying R software (the ecopower package), which allows researchers with pilot data to answer a wide range of design questions to assist them in planning their studies., Comment: 17 pages, 6 figures, 2 code boxes

Deep learning based channel code designs have recently gained interest as an alternative to conventional coding algorithms, particularly for channels for which existing codes do not provide effective solutions. Communication over a feedback channel is one such problem, for which promising results have recently been obtained by employing various deep learning architectures. In this paper, we introduce a novel learning-aided code design for feedback channels, called generalized block attention feedback (GBAF) codes, which i) employs a modular architecture that can be implemented using different neural network architectures; ii) provides order-of-magnitude improvements in the probability of error compared to existing designs; and iii) can transmit at desired code rates.

When working with description logic ontologies, understanding entailments derived by a description logic reasoner is not always straightforward. So far, the standard ontology editor Prot\'eg\'e offers two services to help: (black-box) justifications for OWL 2 DL ontologies, and (glass-box) proofs for lightweight OWL EL ontologies, where the latter exploits the proof facilities of reasoner ELK. Since justifications are often insufficient in explaining inferences, there is thus only little tool support for explaining inferences in more expressive DLs. In this paper, we introduce EVEE-LIBS, a Java library for computing proofs for DLs up to ALCH, and EVEE-PROTEGE, a collection of Prot\'eg\'e plugins for displaying those proofs in Prot\'eg\'e. We also give a short glimpse of the latest version of EVONNE, a more advanced standalone application for displaying and interacting with proofs computed with EVEE-LIBS.

We consider a multicarrier chirp-based waveform for joint radar and communication (JRC) systems and derive its time discrete periodic ambiguity function (AF). An advantage of the waveform is that it includes a set of waveform parameters (e.g., chirp rate) which together with the transmit sequence, can be selected to flexibly shape the AF to be thumbtack-like, or to be ridge-like, either along the delay axis or the Doppler axis. These shapes are applicable for different use cases, e.g., target detection or timeand frequency synchronization. The results show that better signal detection performance than OFDM and DFT-s-OFDM can be achieved on channels with large Doppler frequency. Furthermore, it is shown how transmit sequences can be selected in order to achieve 0 dB peak-to-average-power-ratio (PAPR) of the waveform., Comment: 16 pages, 11 figures, submitted to IEEE Open Journal of the Communications Society

One popular group of defense techniques against adversarial attacks is based on injecting stochastic noise into the network. The main source of robustness of such stochastic defenses however is often due to the obfuscation of the gradients, offering a false sense of security. Since most of the popular adversarial attacks are optimization-based, obfuscated gradients reduce their attacking ability, while the model is still susceptible to stronger or specifically tailored adversarial attacks. Recently, five characteristics have been identified, which are commonly observed when the improvement in robustness is mainly caused by gradient obfuscation. It has since become a trend to use these five characteristics as a sufficient test, to determine whether or not gradient obfuscation is the main source of robustness. However, these characteristics do not perfectly characterize all existing cases of gradient obfuscation, and therefore can not serve as a basis for a conclusive test. In this work, we present a counterexample, showing this test is not sufficient for concluding that gradient obfuscation is not the main cause of improvements in robustness.

Ultra-reliable short-packet communication is a major challenge in future wireless networks with critical applications. To achieve ultra-reliable communications beyond 99.999%, this paper envisions a new interaction-based communication paradigm that exploits feedback from the receiver. We present AttentionCode, a new class of feedback codes leveraging deep learning (DL) technologies. The underpinnings of AttentionCode are three architectural innovations: AttentionNet, input restructuring, and adaptation to fading channels, accompanied by several training methods, including large-batch training, distributed learning, look-ahead optimizer, training-test signal-to-noise ratio (SNR) mismatch, and curriculum learning. The training methods can potentially be generalized to other wireless communication applications with machine learning. Numerical experiments verify that AttentionCode establishes a new state of the art among all DL-based feedback codes in both additive white Gaussian noise (AWGN) channels and fading channels. In AWGN channels with noiseless feedback, for example, AttentionCode achieves a block error rate (BLER) of $10^{-7}$ when the forward channel SNR is 0 dB for a block size of 50 bits, demonstrating the potential of AttentionCode to provide ultra-reliable short-packet communications., Comment: Ultra-reliable short-packet communications, feedback, deep learning, the attention mechanism

Recent studies have proposed that GRAVITY+ instrument is able to trace the circular orbit of the subparsec close-binary supermassive black holes (CB-SMBHs) by measuring the photocentre variation of the hot dust emission. However, the CB-SMBHs orbit may become highly eccentric throughout the evolution of these objects, and the orbital period may be far longer than the observational time baseline. We investigate the problem of detecting the CB-SMBH with hot dust emission and high eccentricity (eCBSMBH, e=0.5) when the observed time baselines of their astrometric data and radial velocities are considerably shorter than the orbital period. The parameter space of the Keplerian model of the eCBSMBH is large for exploratory purposes. We therefore applied the Bayesian method to fit orbital elements of the eCBSMBH to combined radial velocity and astrometric data covering a small fraction of the orbital period. We estimate that a number of potential eCBSMBH systems within reach of GRAVITY + will be similar to the number of the planned circular targets. We show that using observational time baselines that cover ~ 10% of the orbit increases the possibility of determining the period, eccentricity, and total mass of an eCBSMBH. When the observational time baseline becomes too short (~ 5%), the quality of the retrieved eCBSMBH parameters degrades. We also illustrate how interferometry may be used to estimate the photo-centre at the eCBSMBH emission line, which could be relevant for GRAVITY+ successors. Even if the astrometric signal for eCBSMBH systems is reduced by a factor of sqrt{1-e^{2}} compared to circular ones, we find that the hot dust emission of eCBSMBHs can be traced by GRAVITY+ at the elementary level., Comment: accepted for publication in the journal Astronomy & Astrophysics

The state-space models (SSMs) are widely utilized in the analysis of time-series data. SSMs rely on an explicit definition of the state and observation processes. Characterizing these processes is not always easy and becomes a modeling challenge when the dimension of observed data grows or the observed data distribution deviates from the normal distribution. Here, we propose a new formulation of SSM for high-dimensional observation processes. We call this solution the deep direct discriminative decoder (D4). The D4 brings deep neural networks' expressiveness and scalability to the SSM formulation letting us build a novel solution that efficiently estimates the underlying state processes through high-dimensional observation signal. We demonstrate the D4 solutions in simulated and real data such as Lorenz attractors, Langevin dynamics, random walk dynamics, and rat hippocampus spiking neural data and show that the D4 performs better than traditional SSMs and RNNs. The D4 can be applied to a broader class of time-series data where the connection between high-dimensional observation and the underlying latent process is hard to characterize.

We present FREDo, a few-shot document-level relation extraction (FSDLRE) benchmark. As opposed to existing benchmarks which are built on sentence-level relation extraction corpora, we argue that document-level corpora provide more realism, particularly regarding none-of-the-above (NOTA) distributions. Therefore, we propose a set of FSDLRE tasks and construct a benchmark based on two existing supervised learning data sets, DocRED and sciERC. We adapt the state-of-the-art sentence-level method MNAV to the document-level and develop it further for improved domain adaptation. We find FSDLRE to be a challenging setting with interesting new characteristics such as the ability to sample NOTA instances from the support set. The data, code, and trained models are available online (https://github.com/nicpopovic/FREDo)., Comment: Published at NAACL 2022

NASA will launch the Nancy Grace Roman Space Telescope (Roman) in the second half of this decade, which will allow for a generation-defining measurement of dark energy through multiple probes, including Type Ia supernovae (SNe Ia). To improve decisions on survey strategy, we have created the first simulations of realistic Roman images that include artificial SNe Ia injected as point sources in the images. Our analysis combines work done on Roman simulations for weak gravitational lensing studies as well as catalog-level simulations of SN samples. We have created a time series of images over two years containing $\sim$ 1,050 SNe Ia, covering a 1 square degree subarea of a planned 5 square degree deep survey. We have released these images publicly for community use along with input catalogs of all injected sources. We create secondary products from these images by generating coadded images and demonstrating recovery of transient sources using image subtraction. We perform first-use analyses on these images in order to measure galaxy-detection efficiency, point source-detection efficiency, and host-galaxy association biases. The simulated images can be found here: https://roman.ipac.caltech.edu/sims/SN_Survey_Image_sim.html., Comment: 12 pages, 12 figures. Submitted to MNRAS. For simulated images see https://roman.ipac.caltech.edu/sims/SN_Survey_Image_sim.html

In spectra of the Active Galactic Nuclei (AGNs), the [N II] 6548, 6583 A lines are commonly fitted using the fixed intensity ratio of these two lines (R[N II]=I$_{6583}$/I$_{6548}$). However, the used values for fixed intensity ratio are slightly different through literature. There are several theoretical calculations of the transition probabilities which can be used for the line ratio estimation, but there are no experimental measurements of this ratio, since the [N II] lines are extremely weak in laboratory plasma. Therefore, the intensity ratio of [N II] lines can be measured only in the spectra of astrophysical objects. However, precise and systematic measurements have not be done so far, because of difficulties in measurement of the [N II] ratio in various spectra (overlapping with H$\alpha$, weak intensity of [N II], influence of the continuum noise and outflow contribution, etc.). Here we present the measurements of the flux ratio of the [N II]$\lambda\lambda$ 6548, 6583 A emission lines for a sample of 250 Type 2 AGNs spectra taken form Sloan Digital Sky Survey (SDSS) data base. The spectra are chosen to have high signal-to-noise ratio and to [N II] and H$\alpha$ lines do not overlap. The obtained mean flux ratio from measurements is 3.049 $\pm$ 0.021. Our result is in agreement with theoretical result obtained by taking into account the relativistic corrections to the magnetic dipole operator., Comment: Accepted in Advances in Space Research, 8 pages, 4 figures, 1 table

Two-dimensional lead halide perovskites are promising materials for optoelectronics due to the tunability of their properties with the number of lead halide layers and the choice of an organic spacer. Physical understanding for the rational design of materials primarily requires knowledge of crystal structure. 2D lead halide perovskites are usually prepared in the form of films complicating the experimental determination of structure. To enable theoretical studies of experimentally unresolvable structures as well as high-throughput virtual screening, we present an algorithm for crystal structure prediction of lead halide perovskites. Using automatically prepared classical potential we show that our algorithm enables fast access to a structure that can be used for further first-principles studies.

This paper describes and tests a method for carrying out quantified reproducibility assessment (QRA) that is based on concepts and definitions from metrology. QRA produces a single score estimating the degree of reproducibility of a given system and evaluation measure, on the basis of the scores from, and differences between, different reproductions. We test QRA on 18 system and evaluation measure combinations (involving diverse NLP tasks and types of evaluation), for each of which we have the original results and one to seven reproduction results. The proposed QRA method produces degree-of-reproducibility scores that are comparable across multiple reproductions not only of the same, but of different original studies. We find that the proposed method facilitates insights into causes of variation between reproductions, and allows conclusions to be drawn about what changes to system and/or evaluation design might lead to improved reproducibility., Comment: To be published in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL'22)

In most scenarios, conditional image generation can be thought of as an inversion of the image understanding process. Since generic image understanding involves solving multiple tasks, it is natural to aim at generating images via multi-conditioning. However, multi-conditional image generation is a very challenging problem due to the heterogeneity and the sparsity of the (in practice) available conditioning labels. In this work, we propose a novel neural architecture to address the problem of heterogeneity and sparsity of the spatially multi-conditional labels. Our choice of spatial conditioning, such as by semantics and depth, is driven by the promise it holds for better control of the image generation process. The proposed method uses a transformer-like architecture operating pixel-wise, which receives the available labels as input tokens to merge them in a learned homogeneous space of labels. The merged labels are then used for image generation via conditional generative adversarial training. In this process, the sparsity of the labels is handled by simply dropping the input tokens corresponding to the missing labels at the desired locations, thanks to the proposed pixel-wise operating architecture. Our experiments on three benchmark datasets demonstrate the clear superiority of our method over the state-of-the-art and compared baselines. The source code will be made publicly available.