arXiv subjects

Ji Wang

Publications and source records attributed to Ji Wang.

At least 217 records · Page 12Linked to original sources

arXiv2018

Boosting the Robustness Verification of DNN by Identifying the Achilles's Heel

Deep Neural Network (DNN) is a widely used deep learning technique. How to ensure the safety of DNN-based system is a critical problem for the research and application of DNN. Robustness is an important safety property of DNN. However, existing work of verifying DNN's robustness is time-consuming and hard to scale to large-scale DNNs. In this paper, we propose a boosting method for DNN robustness verification, aiming to find counter-examples earlier. Our observation is DNN's different inputs have different possibilities of existing counter-examples around them, and the input with a small difference between the largest output value and the second largest output value tends to be the achilles's heel of the DNN. We have implemented our method and applied it on Reluplex, a state-of-the-art DNN verification tool, and four DNN attacking methods. The results of the extensive experiments on two benchmarks indicate the effectiveness of our boosting method.

cs.LG↗

arXiv2018

Private Model Compression via Knowledge Distillation

The soaring demand for intelligent mobile applications calls for deploying powerful deep neural networks (DNNs) on mobile devices. However, the outstanding performance of DNNs notoriously relies on increasingly complex models, which in turn is associated with an increase in computational expense far surpassing mobile devices' capacity. What is worse, app service providers need to collect and utilize a large volume of users' data, which contain sensitive information, to build the sophisticated DNN models. Directly deploying these models on public mobile devices presents prohibitive privacy risk. To benefit from the on-device deep learning without the capacity and privacy concerns, we design a private model compression framework RONA. Following the knowledge distillation paradigm, we jointly use hint learning, distillation learning, and self learning to train a compact and fast neural network. The knowledge distilled from the cumbersome model is adaptively bounded and carefully perturbed to enforce differential privacy. We further propose an elegant query sample selection method to reduce the number of queries and control the privacy loss. A series of empirical evaluations as well as the implementation on an Android mobile device show that RONA can not only compress cumbersome models efficiently but also provide a strong privacy guarantee. For example, on SVHN, when a meaningful $(9.83,10^{-6})$-differential privacy is guaranteed, the compact model trained by RONA can obtain 20$\times$ compression ratio and 19$\times$ speed-up with merely 0.97% accuracy loss.

cs.LG↗

arXiv2018

Detecting Water In the atmosphere of HR 8799 c with L-band High Dispersion Spectroscopy Aided By Adaptive Optics

High dispersion spectroscopy of brown dwarfs and exoplanets enables exciting science cases, e.g., mapping surface inhomogeneity and measuring spin rate. Here, we present $L$ band observations of HR 8799 c using Keck NIRSPEC (R=15,000) in adaptive optics (AO) mode (NIRSPAO). We search for molecular species (H$_2$O and CH$_4$) in the atmosphere of HR 8799 c with a template matching method, which involves cross correlation between reduced spectrum and a template spectrum. We detect H$_2$O but not CH$_4$, which suggests disequilibrium chemistry in the atmosphere of HR 8799 c, and this is consistent with previous findings. We conduct planet signal injection simulations to estimate the sensitivity of our AO-aided high dispersion spectroscopy observations. We conclude that $10^{-4}$ contrast can be reached in $L$ band. The sensitivity is mainly limited by the accuracy of line list used in modeling spectra and detector noise. The latter will be alleviated by the NIRSPEC upgrade.

astro-ph.EP↗

arXiv2018

Deep exploration of $ε$ Eridani with Keck Ms-band vortex coronagraphy and radial velocities: mass and orbital parameters of the giant exoplanet

We present the most sensitive direct imaging and radial velocity (RV) exploration of $ε$ Eridani to date. $ε$ Eridani is an adolescent planetary system, reminiscent of the early Solar system. It is surrounded by a prominent and complex debris disk which is likely stirred by one or several gas giant exoplanets. The discovery of the RV signature of a giant exoplanet was announced 15 years ago, but has met with scrutiny due to possible confusion with stellar noise. We confirm the planet with a new compilation and analysis of precise RV data spanning 30 years, and combine it with upper limits from our direct imaging search, the most sensitive ever performed. The deep images were taken in the Ms band (4.7$μ$m) with the vortex coronagraph recently installed in W.M. Keck Observatory's infrared camera NIRC2, which opens a sensitive window for planet searches around nearby adolescent systems. The RV data and direct imaging upper limit maps were combined in an innovative joint Bayesian analysis, providing new constraints on the mass and orbital parameters of the elusive planet. $ε$ Eridani b has a mass of $0.78^{+0.38}_{-0.12}$ $M_{Jup}$ and is orbiting $ε$ Eridani at about $3.48\pm 0.02$ AU with a period of $7.37 \pm 0.07$ years. The eccentricity of $ε$ Eridani b's orbit is $0.07^{+0.06}_{-0.05}$, an order of magnitude smaller than early estimates and consistent with a circular orbit. We discuss our findings from the standpoint of planet-disk interactions and prospects for future detection and characterization with the James Webb Space Telescope.

astro-ph.EP↗

arXiv2018

Efficient spectroscopy of exoplanets at small angular separations with vortex fiber nulling

Instrumentation designed to characterize potentially habitable planets may combine adaptive optics and high-resolution spectroscopy techniques to achieve the highest possible sensitivity to spectral signs of life. Detecting the weak signal from a planet containing biomarkers will require exquisite control of the optical wavefront to maximize the planet signal and significantly reduce unwanted starlight. We present an optical technique, known as vortex fiber nulling (VFN), that allows polychromatic light from faint planets at extremely small separations from their host stars ($\lesssimλ/D$) to be efficiently routed to a diffraction-limited spectrograph via a single-mode optical fiber, while light from the star is prevented from entering the spectrograph. VFN takes advantage of the spatial selectivity of a single-mode fiber to isolate the light from close-in companions in a small field of view around the star. We provide theoretical performance predictions of a conceptual design and show that VFN may be utilized to characterize planets detected by radial velocity (RV) instruments in the infrared without knowledge of the azimuthal orientation of their orbits. Using a spectral template-matching technique, we calculate an integration time of $\sim$400, $\sim$100, and $\sim$30 hr for Ross 128 b with Keck, the Thirty Meter Telescope (TMT), and the Large Ultraviolet/Optical/Infrared (LUVOIR) Surveyor, respectively.

astro-ph.IM↗

arXiv2018

Layerwise Perturbation-Based Adversarial Training for Hard Drive Health Degree Prediction

With the development of cloud computing and big data, the reliability of data storage systems becomes increasingly important. Previous researchers have shown that machine learning algorithms based on SMART attributes are effective methods to predict hard drive failures. In this paper, we use SMART attributes to predict hard drive health degrees which are helpful for taking different fault tolerant actions in advance. Given the highly imbalanced SMART datasets, it is a nontrivial work to predict the health degree precisely. The proposed model would encounter overfitting and biased fitting problems if it is trained by the traditional methods. In order to resolve this problem, we propose two strategies to better utilize imbalanced data and improve performance. Firstly, we design a layerwise perturbation-based adversarial training method which can add perturbations to any layers of a neural network to improve the generalization of the network. Secondly, we extend the training method to the semi-supervised settings. Then, it is possible to utilize unlabeled data that have a potential of failure to further improve the performance of the model. Our extensive experiments on two real-world hard drive datasets demonstrate the superiority of the proposed schemes for both supervised and semi-supervised classification. The model trained by the proposed method can correctly predict the hard drive health status 5 and 15 days in advance. Finally, we verify the generality of the proposed training method in other similar anomaly detection tasks where the dataset is imbalanced. The results argue that the proposed methods are applicable to other domains.

cs.LG↗

arXiv2018

Deep Learning Towards Mobile Applications

Recent years have witnessed an explosive growth of mobile devices. Mobile devices are permeating every aspect of our daily lives. With the increasing usage of mobile devices and intelligent applications, there is a soaring demand for mobile applications with machine learning services. Inspired by the tremendous success achieved by deep learning in many machine learning tasks, it becomes a natural trend to push deep learning towards mobile applications. However, there exist many challenges to realize deep learning in mobile applications, including the contradiction between the miniature nature of mobile devices and the resource requirement of deep neural networks, the privacy and security concerns about individuals' data, and so on. To resolve these challenges, during the past few years, great leaps have been made in this area. In this paper, we provide an overview of the current challenges and representative achievements about pushing deep learning on mobile devices from three aspects: training with mobile data, efficient inference on mobile devices, and applications of mobile deep learning. The former two aspects cover the primary tasks of deep learning. Then, we go through our two recent applications that apply the data collected by mobile devices to inferring mood disturbance and user identification. Finally, we conclude this paper with the discussion of the future of this area.

cs.LG↗

arXiv2018

EPIC 203868608: A low-mass quadruple star system in the Upper Scorpius OB association

Young multiple star systems provide excellent testing grounds for theories of star formation and evolution. EPIC 203868608 was previously studied~\citep{David2016} as a triple star system in the Upper Scorpius OB association, but the follow-up Keck NIRC2/HIRES/NIRSPAO observations reported here reveal its quadruple nature. We find that the system consists of a double-lined spectroscopic binary (SB2) Aab (M5+M5) and an eclipsing binary (EB) Bab with a total mass that is lower than that of the SB2. Furthermore, we measure the obliquity of the EB using the Doppler tomography technique during the primary eclipse. EPIC 203868608 Bab is likely on an inclined orbit with a projected obliquity of $-57^{+40}_{-36}$ degrees. The inclined orbit is used to constrain the tidal quality factor for low-mass stars and the evolution of the quadruple system. The analytic framework to infer obliquity that has been developed in this paper can be applied to other EB systems as well as transiting planets.

astro-ph.SR↗

arXiv2018

Direct Imaging in Reflected Light: Characterization of Older, Temperate Exoplanets With 30-m Telescopes

Direct detection, also known as direct imaging, is a method for discovering and characterizing the atmospheres of planets at intermediate and wide separations. It is the only means of obtaining spectra of non-transiting exoplanets. Characterizing the atmospheres of planets in the <5 AU regime, where RV surveys have revealed an abundance of other worlds, requires a 30-m-class aperture in combination with an advanced adaptive optics system, coronagraph, and suite of spectrometers and imagers - this concept underlies planned instruments for both TMT (the Planetary Systems Imager, or PSI) and the GMT (GMagAO-X). These instruments could provide astrometry, photometry, and spectroscopy of an unprecedented sample of rocky planets, ice giants, and gas giants. For the first time habitable zone exoplanets will become accessible to direct imaging, and these instruments have the potential to detect and characterize the innermost regions of nearby M-dwarf planetary systems in reflected light. High-resolution spectroscopy will not only illuminate the physics and chemistry of exo-atmospheres, but may also probe rocky, temperate worlds for signs of life in the form of atmospheric biomarkers (combinations of water, oxygen and other molecular species). By completing the census of non-transiting worlds at a range of separations from their host stars, these instruments will provide the final pieces to the puzzle of planetary demographics. This whitepaper explores the science goals of direct imaging on 30-m telescopes and the technology development needed to achieve them.

astro-ph.IM↗

arXiv2018

The Planetary Systems Imager: 2-5 Micron Channel

We summarize the red channel (2-5 micron) of the Planetary Systems Imager (PSI), a proposed second-generation instrument for the TMT. Cold exoplanets emit the majority of their light in the thermal infrared, which means these exoplanets can be detected at a more modest contrast than at other wavelengths. PSI-Red will be able to detect and characterize a wide variety of exoplanets, including radial-velocity planets on wide orbits, accreting protoplanets in nearby star-forming regions, and reflected-light planets around the nearest stars. PSI-Red will feature an imager, a low-resolution lenslet integral field spectrograph, a medium-resolution lenslet+slicer integral field spectrograph, and a fiber-fed high-resolution spectrograph.

astro-ph.IM↗

arXiv2018

High-contrast imaging of tight resolved binaries with two vector vortex coronagraphs in cascade with the Palomar SDC instrument

More than half of the stars in the solar neighborhood reside in binary/multiple stellar systems, and recent studies suggest that gas giant planets may be more abundant around binaries than single stars. Yet, these multiple systems are usually overlooked or discarded in most direct imaging surveys, as they prove difficult to image at high-contrast using coronographs. This is particularly the case for compact binaries (less than 1" angular separation) with similar stellar magnitudes, where no existing coronagraph can provide high-contrast regime. Here we present preliminary results of an on-going Palomar pilot survey searching for low-mass companions around ~15 young challenging binary systems, with angular separation as close as 0".3 and near-equal K-band magnitudes. We use the Stellar Double Coronagraph (SDC) instrument on the 200-inch Telescope in a modified optical configuration, making it possible to align any targeted binary system behind two vector vortex coronagraphs in cascade. This approach is uniquely possible at Palomar, thanks to the absence of sky rotation combined with the availability of an extreme AO system, and the number of intermediate focal-planes provided by the SDC instrument. Finally, we expose our current data reduction strategy, and we attempt to quantify the exact contrast gain parameter space of our approach, based on our latest observing runs.

astro-ph.IM↗

arXiv2018

Baseline Requirements For Detecting Biosignatures with the HabEx and LUVOIR Mission Concepts

A milestone in understanding life in the universe is the detection of biosignature gases in the atmospheres of habitable exoplanets. Future mission concepts under study by the 2020 decadal survey, e.g., HabEx and LUVOIR, have the potential of achieving this goal. We investigate the baseline requirements for detecting four molecular species, H$_2$O, O$_2$, CH$_4$, and CO$_2$, assuming concentrations of these species equal to that of modern Earth. These molecules are highly relevant to habitability and life on Earth and other planets. Through numerical simulations, we find the minimum requirements of spectral resolution, starlight suppression, and exposure time for detecting biosignature and habitability marker gases. The results are highly dependent on cloud conditions. A low-cloud case is more favorable because of deeper and denser lines whereas a no-cloud case is the pessimistic case for its low albedo. The minimum exposure time for detecting a certain molecule species can vary by a large factor ($\sim$10) between the low-cloud case and the no-cloud case. For all cases, we provide baseline requirements for HabEx and LUVOIR. The impact of exo-zodiacal contamination and thermal background is also discussed and will be included in future studies.

astro-ph.EP↗

arXiv2018

Scheduling Constraint Based Abstraction Refinement for Multi-Threaded Program Verification

Bounded model checking is among the most efficient techniques for the automatic verification of concurrent programs. However, encoding all possible interleavings often requires a huge and complex formula, which significantly limits the salability. This paper proposes a novel and efficient abstraction refinement method for multi-threaded program verification. Observing that the huge formula is usually dominated by the exact encoding of the scheduling constraint, this paper proposes a \tsc based abstraction refinement method, which avoids the huge and complex encoding of BMC. In addition, to obtain an effective refinement, we have devised two graph-based algorithms over event order graph for counterexample validation and refinement generation, which can always obtain a small yet effective refinement constraint. Enhanced by two constraint-based algorithms for counterexample validation and refinement generation, we have proved that our method is sound and complete w.r.t. the given loop unwinding depth. Experimental results on \svcompc benchmarks indicate that our method is promising and significantly outperforms the existing state-of-the-art tools.

cs.PL↗

arXiv2018

Direct demonstration of the emergent magnetism resulting from the multivalence Mn in a LaMnO3 epitaxial thin film system

Atomically engineered oxide heterostructures provide a fertile ground for creating novel states. For example, a two-dimensional electron gas at the interface between two oxide insulators, giant thermoelectric Seebeck coefficient, emergent ferromagnetism from otherwise nonmagnetic components, and colossal ionic conductivity. Extensive research efforts reveal that oxygen deficiency or lattice strain play an important role in determining these unexpected properties. Herein, by studying the abrupt presence of robust ferromagnetism (up to 1.5 uB/Mn) in LaMnO3-based heterostructures, we find the multivalence states of Mn that play a decisive role in the emergence of ferromagnetism in the otherwise antiferromagnetic LaMnO3 thin films. Combining spatially resolved electron energy-loss spectroscopy, X-ray absorption spectroscopy and X-ray magnetic circular dichroism techniques, we determine unambiguously that the ferromagnetism results from a conventional Mn3+-O-Mn4+ double-exchange mechanism rather than an interfacial effect. In contrast, the magnetic dead layer of 5 unit cell in proximity to the interface is found to be accompanied with the accumulation of Mn2+ induced by electronic reconstruction. These findings provide a hitherto-unexplored multivalence state of Mn on the emergent magnetism in undoped manganite epitaxial thin films, such as LaMnO3 and BiMnO3, and shed new light on all-oxide spintronic devices.

cond-mat.mtrl-sci↗

arXiv2018

EarthFinder: A Precise Radial Velocity Probe Mission Concept For the Detection of Earth-Mass Planets Orbiting Sun-like Stars

EarthFinder is a Probe Mission concept selected for study by NASA for input to the 2020 astronomy decadal survey. This study is currently active and a final white paper report is due to NASA at the end of calendar 2018. We are tasked with evaluating the scientific rationale for obtaining precise radial velocity (PRV) measurements in space, which is a two-part inquiry: What can be gained from going to space? What can't be done form the ground? These two questions flow down to these specific tasks for our study - Identify the velocity limit, if any, introduced from micro- and macro-telluric absorption in the Earth's atmosphere; Evaluate the unique advantages that a space-based platform provides to emable the identification and mitigation of stellar acitivity for multi-planet signal recovery.

astro-ph.IM↗

arXiv2017

Searching for the Transit of the Earth--mass exoplanet Proxima~Centauri~b in Antarctica: Preliminary Result

Proxima Centauri is known as the closest star from the Sun. Recently, radial velocity observations revealed the existence of an Earth-mass planet around it. With an orbital period of ~11 days, the surface of Proxima Centauri b is temperate and might be habitable. We took a photometric monitoring campaign to search for its transit, using the Bright Star Survey Telescope at the Zhongshan Station in Antarctica. A transit-like signal appearing on 2016 September 8th, is identified tentatively. Its midtime, $T_{C}=2,457,640.1990\pm0.0017$ HJD, is consistent with the predicted ephemeris based on RV orbit in a 1$σ$ confidence interval. Time-correlated noise is pronounced in the light curve of Proxima Centauri, affecting detection of transits. We develop a technique, in a Gaussian process framework, to gauge the statistical significance of potential transit detection. The tentative transit signal reported here, has a confidence level of $2.5σ$. Further detection of its periodic signals is necessary to confirm the planetary transit of Proxima Centauri b. We plan to monitor Proxima Centauri in next Polar night at Dome A in Antarctica, taking the advantage of continuous darkness. \citet{Kipping17} reported two tentative transit-like signals of Proxima Centauri b, observed by the Microvariability and Oscillation of Stars space Telescope in 2014 and 2015, respectively. The midtransit time of our detection is 138 minutes later than that predicted by their transit ephemeris. If all the signals are real transits, the misalignment of the epochs plausibly suggests transit timing variations of Proxima Centauri b induced by an outer planet in this system.

astro-ph.EP↗

arXiv2017

Untangling Blockchain: A Data Processing View of Blockchain Systems

Blockchain technologies are gaining massive momentum in the last few years. Blockchains are distributed ledgers that enable parties who do not fully trust each other to maintain a set of global states. The parties agree on the existence, values and histories of the states. As the technology landscape is expanding rapidly, it is both important and challenging to have a firm grasp of what the core technologies have to offer, especially with respect to their data processing capabilities. In this paper, we first survey the state of the art, focusing on private blockchains (in which parties are authenticated). We analyze both in-production and research systems in four dimensions: distributed ledger, cryptography, consensus protocol and smart contract. We then present BLOCKBENCH, a benchmarking framework for understanding performance of private blockchains against data processing workloads. We conduct a comprehensive evaluation of three major blockchain systems based on BLOCKBENCH, namely Ethereum, Parity and Hyperledger Fabric. The results demonstrate several trade-offs in the design space, as well as big performance gaps between blockchain and database systems. Drawing from design principles of database systems, we discuss several research directions for bringing blockchain performance closer to the realm of databases.

cs.DB↗

arXiv2017

Generation of higher-order rogue waves from multi-breathers by double degeneracy in an optical fibre

In this paper, we construct a special kind of breather solution of the nonlinear Schrödinger (NLS) equation, the so-called breather-positon ({\it b-positon} for short), which can be obtained by taking the limit $λ_{j}$ $\rightarrow$ $λ_{1}$ of the Lax pair eigenvalues in the order-$n$ periodic solution which is generated by the $n$-fold Darboux transformation from a special "seed" solution--plane wave. Further, an order-$n$ {\it b-positon} gives an order-$n$ rogue wave under a limit $λ_1\rightarrow λ_0$. Here $λ_0$ is a special eigenvalue in a breather of the NLS equation such that its period goes to infinity. Several analytical plots of order-2 breather confirm visually this double degeneration. The last limit in this double degeneration can be realized approximately in an optical fiber governed by the NLS equation, in which an injected initial ideal pulse is created by a frequency comb system and a programable optical filter (wave shaper) according to the profile of an analytical form of the {\it b-positon} at a certain position $z_0$. We also suggest a new way to observe higher-order rogue waves generation in an optical fiber, namely, measure the patterns at the central region of the higher-order {\it b-positon} generated by above ideal initial pulses when $λ_1$ is very close to the $λ_0$. The excellent agreement between the numerical solutions generated from initial ideal inputs with a low signal noise ratio and analytical solutions of order-2 {\it b-positon}, supports strongly this way in a realistic optical fiber system. Our results also show the validity of the generating mechanism of a higher-order rogue waves from a multi-breathers through the double degeneration.

nlin.SI↗