arXiv subjects

Hang Yu

Publications and source records attributed to Hang Yu.

At least 181 records · Page 10Linked to original sources

arXiv2021

Early warning of coalescing neutron-star and neutron-star-black-hole binaries from nonstationary noise background using neural networks

The success of the multi-messenger astronomy relies on gravitational-wave observatories like LIGO and Virgo to provide prompt warning of merger events involving neutron stars (including both binary neutron stars and neutron-star-black-holes), which further depends critically on the low-frequency sensitivity of LIGO as a typical binary neutron star stays in this band for minutes. However, the current sub-60 Hz sensitivity of LIGO has not yet reached its design target and the excess noise can be more than an order of magnitude below 20 Hz. It is limited by nonlinearly coupled noises from auxiliary control loops which are also nonstationary, posing challenges to realistic early-warning pipelines. Nevertheless, machine-learning-based neural networks provide ways to simultaneously improve the low-frequency sensitivity and mitigate its nonstationarity, and detect the real-time gravitational-wave signal with a very short computational time. We propose to achieve this by inputting both the main gravitational-wave readout and key auxiliary witnesses to a compound neural network. Using simulated data with characteristic representing the real LIGO detectors, our machine-learning-based neural networks can reduce nonlinearly coupled noise by about a factor of 5 and allows a typical binary neutron star (neutron-star-black-hole) to be detected 100 s (10 s) before the merger at a distance of 40 Mpc (160 Mpc). If one can further reduce the noise to the fundamental limit, our neural networks can achieve detection out to a distance of 80 Mpc and 240 Mpc for binary neutron stars and neutron-star-black-holes, respectively. It thus demonstrates that utilizing machine-learning-based neural networks is a promising direction for the timely detection of the coalescence of electromagnetically bright LIGO/Virgo sources.

gr-qc↗

arXiv2021

Tides in the high-eccentricity migration of hot Jupiters: Triggering diffusive growth by nonlinear mode interactions

High eccentricity migration is a possible formation channel for hot Jupiters. However, in order for it to be consistent with the observed population of planets, tides must circularize the orbits in less than $\approx$ a Myr. A potential mechanism for such rapid circularization is the diffusive growth of the tidally driven planetary f-mode. Such growth occurs if the f-mode's phase at pericenter varies chaotically from one pericenter passage to the next. Previous studies focused on the variation of the orbital period due to tidal back-reaction on the orbit as the source of chaos. Here we show that nonlinear mode interactions can also be an important source. Specifically, we show that nonlinear interactions between a parent f-mode and daughter f-/p-modes induce an energy-dependent shift in the oscillation frequency of the parent. This frequency shift varies randomly from orbit to orbit because the parent's energy varies. As a result, the parent's phase at pericenter varies randomly, which we find can trigger it to grow diffusively. We show that the phase shift induced by nonlinear mode interactions in fact dominates the shift induced by tidal back-reaction and significantly lowers the one-kick energy threshold for diffusive growth by about a factor of 5 compared to the linear theory's prediction. Nonlinear interactions could thus enhance the formation rate of hot Jupiters through the high-eccentricity migration channel and potentially mitigate the discrepancy between the observed and predicted occurrence rates for close-in gas giants as compared to those further from the star.

astro-ph.EP↗

arXiv2021

A General Framework of Nonparametric Feature Selection in High-Dimensional Data

Nonparametric feature selection in high-dimensional data is an important and challenging problem in statistics and machine learning fields. Most of the existing methods for feature selection focus on parametric or additive models which may suffer from model misspecification. In this paper, we propose a new framework to perform nonparametric feature selection for both regression and classification problems. In this framework, we learn prediction functions through empirical risk minimization over a reproducing kernel Hilbert space. The space is generated by a novel tensor product kernel which depends on a set of parameters that determine the importance of the features. Computationally, we minimize the empirical risk with a penalty to estimate the prediction and kernel parameters at the same time. The solution can be obtained by iteratively solving convex optimization problems. We study the theoretical property of the kernel feature space and prove both the oracle selection property and the Fisher consistency of our proposed method. Finally, we demonstrate the superior performance of our approach compared to existing methods via extensive simulation studies and application to a microarray study of eye disease in animals.

stat.ME↗

arXiv2021

Detecting resonant tidal excitations of Rossby modes in coalescing neutron-star binaries with third-generation gravitational-wave detectors

Rossby modes (r-modes) of rotating neutron stars can be excited by the gravitomagnetic forces in coalescing binary systems. The previous study by Flanagan and Racine [Phys. Rev. D 75, 044001 (2007)] showed that this kind of dynamical tide (DT) can induce phase shifts of 0.1 rad on gravitational waveforms, which is detectable by third-generation (3G) detectors. In this paper, we study the impact of this DT on measuring neutron-star parameters in the era of 3G detectors. We incorporate two universal relations among neutron star properties predicted by different equations of state: (i) the well-known I-Love relation between momentum of inertia and (f-mode) tidal Love number, and (ii) a relation between the r-mode overlap and tidal Love number, which is newly explored in this paper. We find that r-mode DT will provide rich information about slowly rotating neutron stars with frequency ranging from 10 to 100 Hz. For a binary neutron star system (with a signal-to-noise ratio around 1500 in the Cosmic Explorer), the spin frequency of each individual neutron star can be constrained to 6% (fractional error) in the best-case scenario. The degeneracy between the Love numbers of individual neutron stars is dramatically reduced: each individual Love number can be constrained to around 20% in the best case, while the fractional error for both symmetric and anti-symmetric Love numbers are reduced by factors of around 300. Furthermore, DT also allows us to measure the spin inclination angles of the neutron stars, to 0.09 rad in the best case, and thus place constraints on NS natal kicks and supernova explosion models. Besides parameter estimation, we have also developed a semi-analytic method that accurately describes detailed features of the binary evolution that arise due to the DT.

gr-qc↗

arXiv2021

A unified construction of all-speed HLL-type schemes for hypersonic heating computations

In this paper, a unified framework to develop all-speed HLL-type schemes for hypersonic heating computations is constructed. Such a unified construction method combines two effective improving techniques: a shock robustness improvement and a low-Mach number fix. It is implemented by properly modifying the approximate solutions of the local Riemann problem in the HLL framework, resulting in two all-speed HLL-type schemes, namely ASHLLC and ASHLLEM solvers. Results from both numerical analysis and experiments demonstrate that the newly proposed schemes not only preserve desirable properties of their original versions, but are also able to provide accurate and robust solutions for complex flows ranging from low-Mach number incompressible to hypersonic compressible regimes. Thus, both the ASHLLC and ASHLLEM schemes can be used as reliable methods for hypersonic heating computations.

cs.CE↗

arXiv2021

Further studies on numerical instabilities of Godunov-type schemes for strong shocks

In this paper, continuous research is undertaken to explore the underlying mechanism of numerical shock instabilities of Godunov-type schemes for strong shocks. By conducting dissipation analysis of Godunov-type schemes and a sequence of numerical experiments, we are able to clarify that the instability may be attributed to insufficient entropy production inside the numerical shock structure. As a result, a general entropy-control technique for improving the robustness of various Godunov-type schemes at strong shocks is developed. It plays a part in guaranteeing that enough entropy is produced inside the numerical shock structure. Furthermore, such a modified approach does not introduce any additional numerical dissipation on linear degenerate waves to suppress the shock instability. Numerical results that are obtained for various test cases indicate that the proposed methods have a good performance in terms of accuracy and robustness.

physics.comp-ph↗

arXiv2020

Elastic Net based Feature Ranking and Selection

Feature selection is important in data representation and intelligent diagnosis. Elastic net is one of the most widely used feature selectors. However, the features selected are dependant on the training data, and their weights dedicated for regularized regression are irrelevant to their importance if used for feature ranking, that degrades the model interpretability and extension. In this study, an intuitive idea is put at the end of multiple times of data splitting and elastic net based feature selection. It concerns the frequency of selected features and uses the frequency as an indicator of feature importance. After features are sorted according to their frequency, linear support vector machine performs the classification in an incremental manner. At last, a compact subset of discriminative features is selected by comparing the prediction performance. Experimental results on breast cancer data sets (BCDR-F03, WDBC, GSE 10810, and GSE 15852) suggest that the proposed framework achieves competitive or superior performance to elastic net and with consistent selection of fewer features. How to further enhance its consistency on high-dimension small-sample-size data sets should be paid more attention in our future work. The proposed framework is accessible online (https://github.com/NicoYuCN/elasticnetFR).

cs.LG↗

arXiv2020

Training Robust Deep Neural Networks via Adversarial Noise Propagation

In practice, deep neural networks have been found to be vulnerable to various types of noise, such as adversarial examples and corruption. Various adversarial defense methods have accordingly been developed to improve adversarial robustness for deep models. However, simply training on data mixed with adversarial examples, most of these models still fail to defend against the generalized types of noise. Motivated by the fact that hidden layers play a highly important role in maintaining a robust model, this paper proposes a simple yet powerful training algorithm, named \emph{Adversarial Noise Propagation} (ANP), which injects noise into the hidden layers in a layer-wise manner. ANP can be implemented efficiently by exploiting the nature of the backward-forward training style. Through thorough investigations, we determine that different hidden layers make different contributions to model robustness and clean accuracy, while shallow layers are comparatively more critical than deep layers. Moreover, our framework can be easily combined with other adversarial training methods to further improve model robustness by exploiting the potential of hidden layers. Extensive experiments on MNIST, CIFAR-10, CIFAR-10-C, CIFAR-10-P, and ImageNet demonstrate that ANP enables the strong robustness for deep models against both adversarial and corrupted ones, and also significantly outperforms various adversarial defense methods.

cs.LG↗

arXiv2020

Direct determination of supermassive black hole properties with gravitational-wave radiation from surrounding stellar-mass black hole binaries

A significant number of stellar-mass black-hole (BH) binaries may merge in galactic nuclei or in the surrounding gas disks. With purposed space-borne gravitational-wave observatories, we may use such a binary as a signal carrier to probe modulations induced by a central supermassive BH (SMBH), which further allows us to place constraints on the SMBH's properties. We show in particular the de Sitter precession of the inner stellar-mass binary's orbital angular momentum (AM) around the AM of the outer orbit will be detectable if the precession period is comparable to the duration of observation, typically a few years. Once detected, the precession can be combined with the Doppler shift arising from the outer orbital motion to determine the mass of the SMBH and the outer orbital separation individually and each with percent-level accuracy. If we further assume a joint detection by space-borne and ground-based detectors, the detectability threshold could be extended to a precession period of ~100 yr.

gr-qc↗

arXiv2020

Interpreting and Improving Adversarial Robustness of Deep Neural Networks with Neuron Sensitivity

Deep neural networks (DNNs) are vulnerable to adversarial examples where inputs with imperceptible perturbations mislead DNNs to incorrect results. Despite the potential risk they bring, adversarial examples are also valuable for providing insights into the weakness and blind-spots of DNNs. Thus, the interpretability of a DNN in the adversarial setting aims to explain the rationale behind its decision-making process and makes deeper understanding which results in better practical applications. To address this issue, we try to explain adversarial robustness for deep models from a new perspective of neuron sensitivity which is measured by neuron behavior variation intensity against benign and adversarial examples. In this paper, we first draw the close connection between adversarial robustness and neuron sensitivities, as sensitive neurons make the most non-trivial contributions to model predictions in the adversarial setting. Based on that, we further propose to improve adversarial robustness by constraining the similarities of sensitive neurons between benign and adversarial examples which stabilizes the behaviors of sensitive neurons towards adversarial noises. Moreover, we demonstrate that state-of-the-art adversarial training methods improve model robustness by reducing neuron sensitivities which in turn confirms the strong connections between adversarial robustness and neuron sensitivity as well as the effectiveness of using sensitive neurons to build robust models. Extensive experiments on various datasets demonstrate that our algorithm effectively achieves excellent results.

cs.CV↗

arXiv2020

Spin and Eccentricity Evolution in Triple Systems: from the Lidov-Kozai Interaction to the Final Merger of the Inner Binary

We study the spin and eccentricity evolution of black-hole (BH) binaries that are perturbed by tertiary masses and experience the Lidov-Kozai (LK) excitation. We focus on three aspects. Firstly, we study the spin-orbit alignment of the inner binary following the approach outlined by Antonini et al. [MNRAS 480, L58 (2018)] and Liu and Lai [ApJ 863, 68 (2018)], yet allowing the spins to have random initial orientations. We confirm the existence of a dynamical attractor that drives the spin-orbit angle at the end of the LK evolution to a value given by the initial angle between the spin and the outer orbital angular momentum (instead of to a specific value of the effective spin). Secondly, we follow the (inner) binary's evolution further to the merger to study the final spin-spin alignment. We generalize the effective potential theory to include orbital eccentricity, which allows us to efficiently evolve the system in the early inspiral stages. We further find that the spin-spin and spin-orbit alignments are correlated and the correlation is determined by the initial spin-orbit angle. For systems with the spin vectors initially in the orbital plane, the final spins strongly disfavor an aligned configuration and could thus lead to a greater value of the GW recoil than a uniform spin-spin alignment would predict. Lastly, we study the maximum eccentricity excitation that can be achieved during the LK process, including the effects of gravitational-wave radiation. We find that when the tertiary mass is a super-massive BH and the inner binary is massive, then even with the maximum LK excitation, the residual eccentricity is typically less than 0.1 when the binary's orbital frequency reaches 10 Hz, and a decihertz detector would be necessary to follow such a system's orbital evolution.

gr-qc↗

arXiv2020

Tidally excited oscillations in hot white dwarfs

We study the flux variation in helium white dwarfs (WDs) induced by dynamical tides for a variety of WD models with effective temperatures ranging from $T$=10 kK to $T$=26 kK. At linear order, we find the dynamical tide can significantly perturb the observed flux in hot WDs. If the temperature $T\gtrsim14$ kK, then the dynamical tide may induce a fractional change in the flux by >1% when the orbital period is $P_{\rm orb}\simeq 20-60\,{\rm min}$. The ratio between the flux modulation due to the dynamical tide and that due to the equilibrium tide (i.e., ellipsoidal variability) increases as the WD's radius decreases, and it could exceed O(10) if the WD has a radius $R\lesssim0.03 R_\odot$. Unlike the ellipsoidal variability which is in phase with the orbital motion, the pulsation caused by the dynamical tide may have a substantial phase shift. A cold WD with $T\lesssim 10$ kK, on the other hand, is unlikely to show observable pulsations due to the dynamical tide. At shorter orbital periods, the dynamical tide may become highly nonlinear. We approximate this regime by treating the waves as one-way traveling waves and find the flux variation is typically reduced to 0.1%-1% and the excess phase is likely to be 90 degrees (though with large uncertainty). Even in the traveling-wave limit, the flux perturbation due to dynamical tide could still exceed the ellipsoidal variability for compact WDs with $R\lesssim0.02 R_\odot$. We further estimate the nonlinear flux perturbations oscillating at four times the orbital frequency dominated by a self-coupled parent g-mode driving low-order daughter p-modes. The nonlinear flux variation could be nearly 50% of the linear variation for very hot WD models with $T\gtrsim26$ kK and 1% linear flux variation. We thus predict both the linear and nonlinear flux variations due to dynamical tides are likely to have significant observational signatures.

astro-ph.SR↗

arXiv2020

Hunting for Dark Matter Subhalos in Strong Gravitational Lensing with Neural Networks

Dark matter substructures are interesting since they can reveal the properties of dark matter. Collisionless N-body simulations of cold dark matter show more substructures compared with the population of dwarf galaxy satellites observed in our local group. Therefore, understanding the population and property of subhalos at cosmological scale would be an interesting test for cold dark matter. In recent years, it has become possible to detect individual dark matter subhalos near images of strongly lensed extended background galaxies. In this work, we discuss the possibility of using deep neural networks to detect dark matter subhalos, and showing some preliminary results with simulated data. We found that neural networks not only show promising results on detecting multiple dark matter subhalos, but also learn to reject the subhalos on the lensing arc of a smooth lens where there is no subhalo.

astro-ph.CO↗

arXiv2020

A task-based approach to parallel parametric linear programming solving, and application to polyhedral computations

Parametric linear programming is a central operation for polyhedral computations, as well as in certain control applications.Here we propose a task-based scheme for parallelizing it, with quasi-linear speedup over large problems.This type of parallel applications is challenging, because several tasks mightbe computing the same region. In this paper, we are presenting thealgorithm itself with a parallel redundancy elimination algorithm, andconducting a thorough performance analysis.

cs.CG↗

arXiv2020

Context Model for Pedestrian Intention Prediction using Factored Latent-Dynamic Conditional Random Fields

Smooth handling of pedestrian interactions is a key requirement for Autonomous Vehicles (AV) and Advanced Driver Assistance Systems (ADAS). Such systems call for early and accurate prediction of a pedestrian's crossing/not-crossing behaviour in front of the vehicle. Existing approaches to pedestrian behaviour prediction make use of pedestrian motion, his/her location in a scene and static context variables such as traffic lights, zebra crossings etc. We stress on the necessity of early prediction for smooth operation of such systems. We introduce the influence of vehicle interactions on pedestrian intention for this purpose. In this paper, we show a discernible advance in prediction time aided by the inclusion of such vehicle interaction context. We apply our methods to two different datasets, one in-house collected - NTU dataset and another public real-life benchmark - JAAD dataset. We also propose a generic graphical model Factored Latent-Dynamic Conditional Random Fields (FLDCRF) for single and multi-label sequence prediction as well as joint interaction modeling tasks. FLDCRF outperforms Long Short-Term Memory (LSTM) networks across the datasets ($\sim$100 sequences per dataset) over identical time-series features. While the existing best system predicts pedestrian stopping behaviour with 70\% accuracy 0.38 seconds before the actual events, our system achieves such accuracy at least 0.9 seconds on an average before the actual events across datasets.

cs.CV↗

arXiv2020

Bias-based Universal Adversarial Patch Attack for Automatic Check-out

Adversarial examples are inputs with imperceptible perturbations that easily misleading deep neural networks(DNNs). Recently, adversarial patch, with noise confined to a small and localized patch, has emerged for its easy feasibility in real-world scenarios. However, existing strategies failed to generate adversarial patches with strong generalization ability. In other words, the adversarial patches were input-specific and failed to attack images from all classes, especially unseen ones during training. To address the problem, this paper proposes a bias-based framework to generate class-agnostic universal adversarial patches with strong generalization ability, which exploits both the perceptual and semantic bias of models. Regarding the perceptual bias, since DNNs are strongly biased towards textures, we exploit the hard examples which convey strong model uncertainties and extract a textural patch prior from them by adopting the style similarities. The patch prior is more close to decision boundaries and would promote attacks. To further alleviate the heavy dependency on large amounts of data in training universal attacks, we further exploit the semantic bias. As the class-wise preference, prototypes are introduced and pursued by maximizing the multi-class margin to help universal training. Taking AutomaticCheck-out (ACO) as the typical scenario, extensive experiments including white-box and black-box settings in both digital-world(RPC, the largest ACO related dataset) and physical-world scenario(Taobao and JD, the world' s largest online shopping platforms) are conducted. Experimental results demonstrate that our proposed framework outperforms state-of-the-art adversarial patch attack methods.

cs.CV↗

arXiv2020

Astrophysics and cosmology with a decihertz gravitational-wave detector: TianGO

We present the astrophysical science case for a space-based, decihertz gravitational-wave (GW) detector. We particularly highlight an ability to infer a source's sky location, both when combined with a network of ground-based detectors to form a long triangulation baseline, and by itself for the early warning of merger events. Such an accurate location measurement is the key for using GW signals as standard sirens for constraining the Hubble constant. This kind of detector also opens up the possibility to test type Ia supernovae progenitor hypotheses by constraining the merger rates of white dwarf binaries with both super- and sub-Chandrasekhar masses separately. We will discuss other scientific outcomes that can be delivered, including the constraint of structure formation in the early Universe, the search for intermediate-mass black holes, the precise determination of black hole spins, the probe of binary systems' orbital eccentricity evolution, and the detection of tertiary masses around merging binaries.

gr-qc↗

arXiv2020

Excitation of f-modes during mergers of spinning binary neutron star

Tidal effects have important imprints on gravitational waves (GWs) emitted during the final stage of the coalescence of binaries that involve neutron stars (NSs). Dynamical tides can be significant when NS oscillations become resonant with orbital motion; understanding this process is important for accurately modeling GW emission from these binaries, and for extracting NS information from GW data. In this paper, we carry out a systematic study on the tidal excitation of fundamental modes of spinning NSs in coalescencing binaries, focusing on the case when the NS spin is anti-aligned with the orbital angular momentum-where the tidal resonance is most likely to take place. We first expand NS oscillations into stellar eigen-modes, and then obtain a Hamiltonian that governs the tidally coupled orbit-mode evolution. We next find a new approximation that can lead to analytic expressions of tidal excitations to a high accuracy, and are valid in all regimes of the binary evolution: adiabatic, resonant, and post-resonance. Using the method of osculating orbits, we obtain semi-analytic approximations to the orbital evolution and GW emission; their agreements with numerical results give us confidence in on our understanding of the system's dynamics. In particular, we recover both the averaged post-resonance evolution, which differs from the pre-resonance point-particle orbit by shifts in orbital energy and angular momentum, as well as instantaneous perturbations driven by the tidal motion. Finally, we use the Fisher matrix technique to study the effect of dynamical tides on parameter estimation. We find that the dynamical tides may potentially provide an additional channel to study the physics of NSs. The method presented in this paper is generic and not restricted to f mode; it can also be applied to other types of tide.

gr-qc↗