Searcharxiv⌕ Search

arXiv subjects

Xi Zhang

Publications and source records attributed to Xi Zhang.

At least 307 records · Page 17Linked to original sources
arXiv2020

MDLdroid: a ChainSGD-reduce Approach to Mobile Deep Learning for Personal Mobile Sensing

Personal mobile sensing is fast permeating our daily lives to enable activity monitoring, healthcare and rehabilitation. Combined with deep learning, these applications have achieved significant success in recent years. Different from conventional cloud-based paradigms, running deep learning on devices offers several advantages including data privacy preservation and low-latency response for both model inference and update. Since data collection is costly in reality, Google's Federated Learning offers not only complete data privacy but also better model robustness based on multiple user data. However, personal mobile sensing applications are mostly user-specific and highly affected by environment. As a result, continuous local changes may seriously affect the performance of a global model generated by Federated Learning. In addition, deploying Federated Learning on a local server, e.g., edge server, may quickly reach the bottleneck due to resource constraint and serious failure by attacks. Towards pushing deep learning on devices, we present MDLdroid, a novel decentralized mobile deep learning framework to enable resource-aware on-device collaborative learning for personal mobile sensing applications. To address resource limitation, we propose a ChainSGD-reduce approach which includes a novel chain-directed Synchronous Stochastic Gradient Descent algorithm to effectively reduce overhead among multiple devices. We also design an agent-based multi-goal reinforcement learning mechanism to balance resources in a fair and efficient manner. Our evaluations show that our model training on off-the-shelf mobile devices achieves 2x to 3.5x faster than single-device training, and 1.5x faster than the master-slave approach.

cs.LG↗
arXiv2020

Ultra High Fidelity Image Compression with $\ell_\infty$-constrained Encoding and Deep Decoding

In many professional fields, such as medicine, remote sensing and sciences, users often demand image compression methods to be mathematically lossless. But lossless image coding has a rather low compression ratio (around 2:1 for natural images). The only known technique to achieve significant compression while meeting the stringent fidelity requirements is the methodology of $\ell_\infty$-constrained coding that was developed and standardized in nineties. We make a major progress in $\ell_\infty$-constrained image coding after two decades, by developing a novel CNN-based soft $\ell_\infty$-constrained decoding method. The new method repairs compression defects by using a restoration CNN called the $\ell_\infty\mbox{-SDNet}$ to map a conventionally decoded image to the latent image. A unique strength of the $\ell_\infty\mbox{-SDNet}$ is its ability to enforce a tight error bound on a per pixel basis. As such, no small distinctive structures of the original image can be dropped or distorted, even if they are statistical outliers that are otherwise sacrificed by mainstream CNN restoration methods. More importantly, this research ushers in a new image compression system of $\ell_\infty$-constrained encoding and deep soft decoding ($\ell_\infty\mbox{-ED}^2$). The $\ell_\infty \mbox{-ED}^2$ approach beats the best of existing lossy image compression methods (e.g., BPG, WebP, etc.) not only in $\ell_\infty$ but also in $\ell_2$ error metric and perceptual quality, for bit rates near the threshold of perceptually transparent reconstruction. Operationally, the new compression system is practical, with a low-complexity real-time encoder and a cascade decoder consisting of a fast initial decoder and an optional CNN soft decoder.

eess.IV↗
arXiv2020

Deflating Super-Puffs: Impact of Photochemical Hazes on the Observed Mass-Radius Relationship of Low Mass Planets

The observed mass-radius relationship of low-mass planets informs our understanding of their composition and evolution. Recent discoveries of low mass, large radii objects ("super-puffs") have challenged theories of planet formation and atmospheric loss, as their high inferred gas masses make them vulnerable to runaway accretion and hydrodynamic escape. Here we propose that high altitude photochemical hazes could enhance the observed radii of low-mass planets and explain the nature of super-puffs. We construct model atmospheres in radiative-convective equilibrium and compute rates of atmospheric escape and haze distributions, taking into account haze coagulation, sedimentation, diffusion, and advection by an outflow wind. We develop mass-radius diagrams that include atmospheric lifetimes and haze opacity, which is enhanced by the outflow, such that young (~0.1-1 Gyr), warm (T$_{eq}$ $\geq$ 500 K), low mass objects ($M_c$ < 4M$_{\rm Earth}$) should experience the most apparent radius enhancement due to hazes, reaching factors of three. This reconciles the densities and ages of the most extreme super-puffs. For Kepler-51b, the inclusion of hazes reduces its inferred gas mass fraction to <10%, similar to that of planets on the large radius side of the sub-Neptune radius gap. This suggests that Kepler-51b may be evolving towards that population, and that some warm sub-Neptunes may have evolved from super-puffs. Hazes also render transmission spectra of super-puffs and sub-Neptunes featureless, consistent with recent measurements. Our hypothesis can be tested by future observations of super-puffs' transmission spectra at mid-infrared wavelengths, where we predict that the planet radius will be half of that observed in the near-infrared.

astro-ph.EP↗
arXiv2020

Near-lossless $\ell_\infty$-constrained Image Decompression via Deep Neural Network

Recently a number of CNN-based techniques were proposed to remove image compression artifacts. As in other restoration applications, these techniques all learn a mapping from decompressed patches to the original counterparts under the ubiquitous $\ell_\infty$ metric. However, this approach is incapable of restoring distinctive image details which may be statistical outliers but have high semantic importance (e.g., tiny lesions in medical images). To overcome this weakness, we propose to incorporate an $\ell_\infty$ fidelity criterion in the design of neural network so that no small, distinctive structures of the original image can be dropped or distorted. Experimental results demonstrate that the proposed method outperforms the state-of-the-art methods in $\ell_\infty$ error metric and perceptual quality, while being competitive in $\ell_2$ error metric as well. It can restore subtle image details that are otherwise destroyed or missed by other algorithms. Our research suggests a new machine learning paradigm of ultra high fidelity image compression that is ideally suited for applications in medicine, space, and sciences.

cs.CV↗
arXiv2020

Synthesis and temperature-dependent photoluminescence of high density GeSe triangular nanoplate arrays on Si substrates

We have grown germanium selenide (GeSe) triangular nanoplate arrays (TNAs) with a high density (3.82E+6 / mm2) on the Si (111) substrate using a simple thermal evaporation method. The thickness and trilateral lengths of a single triangular nanoplate were statistically estimated by atomic force microscopy (AFM) as 44 nm, 365 nm, 458 nm and 605 nm, respectively. Transmission electron microscopy (TEM) images and X-ray diffraction (XRD) patterns show that the TNAs were composed of single crystalline GeSe phase. The Se-related defects in the lattice were also revealed by TEM images and Raman vibration modes. Unlike previously reported GeSe compounds, the GeSe TNAs exhibited temperature-dependent photoluminescence (PL). In addition, not previously reported PL peak (1.25 eV) of the 44 nm thick TNAs at 5 K was in the gaps between those of GeSe monolayers (1.5 nm) and thin films (400 nm), revealing a close relationship between the PL peak and the thickness of GeSe. The high-density structure and temperature-dependent PL of the TNAs on the Si substrate may be useful for temperature controllable semiconductor nanodevices.

cond-mat.mes-hall↗
arXiv2020

Rapid Synthesis of Thermoelectric SnSe Thin Films by MPCVD

Microwave plasma chemical vapor deposition (MPCVD) has been traditionally used to synthesize carbon-based materials such as diamonds, carbon nanotubes and graphene. Here we report that a rapid and catalyst-free growth of SnSe thin films can be achieved by using single-mode MPCVD with appropriate source materials. The analysis combing microscope images, X-ray diffraction patterns and lattice vibration modes shows that the grown thin films were composed of orthorhombic structured SnSe polycrystals mainly along the (111) direction. Further thermoelectric (TE) characterizations reveal that the power factor of the SnSe films reached 3.98 μW cm-1K-2 at 600 K, comparable to the highest reported values of SnSe thin films. Our results may open an avenue for rapid synthesis of new types of materials such as IV-VI compounds and be useful for TE application of these materials.

cond-mat.mtrl-sci↗
arXiv2019

Finding Route Hotspots in Large Labeled Networks

In many advanced network analysis applications, like social networks, e-commerce, and network security, hotspots are generally considered as a group of vertices that are tightly connected owing to the similar characteristics, such as common habits and location proximity. In this paper, we investigate the formation of hotspots from an alternative perspective that considers the routes along the network paths as the auxiliary information, and attempt to find the route hotspots in large labeled networks. A route hotspot is a cohesive subgraph that is covered by a set of routes, and these routes correspond to the same sequential pattern consisting of vertices' labels. To the best of our knowledge, the problem of Finding Route Hotspots in Large Labeled Networks has not been tackled in the literature. However, it is challenging as counting the number of hotspots in a network is #P-hard. Inspired by the observation that the sizes of hotspots decrease with the increasing lengths of patterns, we prove several anti-monotonicity properties of hotspots, and then develop a scalable algorithm called FastRH that can use these properties to effectively prune the patterns that cannot form any hotspots. In addition, to avoid the duplicate computation overhead, we judiciously design an effective index structure called RH-Index for storing the hotspot and pattern information collectively, which also enables incremental updating and efficient query processing. Our experimental results on real-world datasets clearly demonstrate the effectiveness and scalability of our proposed methods.

cs.DS↗
arXiv2019

Transit Signatures of Inhomogeneous Clouds on Hot Jupiters: Insights From Microphysical Cloud Modeling

We determine the observability in transmission of inhomogeneous cloud cover on the limbs of hot Jupiters through post processing a general circulation model to include cloud distributions computed using a cloud microphysics model. We find that both the east and west limb often form clouds, but that the different properties of these clouds enhances the limb to limb differences compared to the clear case. Using JWST it should be possible to detect the presence of cloud inhomogeneities by comparing the shape of the transit lightcurve at multiple wavelengths because inhomogeneous clouds impart a characteristic, wavelength dependent signature. This method is statistically robust even with limited wavelength coverage, uncertainty on limb darkening coefficients, and imprecise transit times. We predict that the short wavelength slope varies strongly with temperature. The hot limb of the hottest planets form higher altitude clouds composed of smaller particles leading to a strong rayleigh slope. The near infrared spectral features of clouds are almost always detectable, even when no spectral slope is visible in the optical. In some of our models a spectral window between 5 and 9 microns can be used to probe through the clouds and detect chemical spectral features. Our cloud particle size distributions are not log-normal and differ from species to species. Using the area or mass weighted particle size significantly alters the relative strength of the cloud spectral features compared to using the predicted size distribution. Finally, the cloud content of a given planet is sensitive to a species' desorption energy and contact angle, two parameters that could be constrained experimentally in the future.

astro-ph.EP↗
arXiv2019

Scalable Explanation of Inferences on Large Graphs

Probabilistic inferences distill knowledge from graphs to aid human make important decisions. Due to the inherent uncertainty in the model and the complexity of the knowledge, it is desirable to help the end-users understand the inference outcomes. Different from deep or high-dimensional parametric models, the lack of interpretability in graphical models is due to the cyclic and long-range dependencies and the byzantine inference procedures. Prior works did not tackle cycles and make \textit{the} inferences interpretable. To close the gap, we formulate the problem of explaining probabilistic inferences as a constrained cross-entropy minimization problem to find simple subgraphs that faithfully approximate the inferences to be explained. We prove that the optimization is NP-hard, while the objective is not monotonic and submodular to guarantee efficient greedy approximation. We propose a general beam search algorithm to find simple trees to enhance the interpretability and diversity in the explanations, with parallelization and a pruning strategy to allow efficient search on large and dense graphs without hurting faithfulness. We demonstrate superior performance on 10 networks from 4 distinct applications, comparing favorably to other explanation methods. Regarding the usability of the explanation, we visualize the explanation in an interface that allows the end-users to explore the diverse search results and find more personalized and sensible explanations.

cs.SI↗
arXiv2019

Atmospheric Circulation of Brown Dwarfs and Jupiter and Saturn-like Planets: Zonal Jets, Long-term Variability, and QBO-type Oscillations

Brown dwarfs and directly imaged giant planets exhibit significant evidence for active atmospheric circulation, which induces a large-scale patchiness in the cloud structure that evolves significantly over time, as evidenced by infrared light curves and Doppler maps. These observations raise critical questions about the fundamental nature of the circulation, its time variability, and the overall relationship to the circulation on Jupiter and Saturn. Jupiter and Saturn themselves exhibit numerous robust zonal (east-west) jet streams at the cloud level; moreover, both planets exhibit long-term stratospheric oscillations involving perturbations of zonal wind and temperature that propagate downward over time on timescales of ~4 years (Jupiter) and ~15 years (Saturn). These oscillations, dubbed the Quasi Quadrennial Oscillation (QQO) for Jupiter and the Semi-Annual Oscillation (SAO) on Saturn, are thought to be analogous to the Quasi-Biennial Oscillation (QBO) on Earth, which is driven by upward propagation of equatorial waves from the troposphere. To investigate these issues, we here present global, three-dimensional, high-resolution numerical simulations of the flow in the stratified atmosphere--overlying the convective interior--of brown dwarfs and Jupiter-like planets. The effect of interior convection is parameterized by inducing small-scale, randomly varying perturbations in the radiative-convective boundary at the base of the model. In the simulations, the convective perturbations generate atmospheric waves and turbulence that interact with the rotation to produce numerous zonal jets. Moreover, the equatorial stratosphere exhibits stacked eastward and westward jets that migrate downward over time, exactly as occurs in the terrestrial QBO, Jovian QQO, and Saturnian SAO. This is the first demonstration of a QBO-like phenomenon in 3D numerical simulations of a giant planet.

astro-ph.EP↗
arXiv2019

Analytically stable Higgs bundles on some non-Kähler manifolds

In this paper, we study Higgs bundles on non-compact Hermitian manifolds. Under some assumptions for the underlying Hermitian manifolds which are not necessarily Kähler, we solve the Hermitian-Einstein equation on analytically stable Higgs bundles.

math.DG↗
arXiv2019

When Full Duplex Wireless Meets Non-Orthogonal Multiple Access: Opportunities and Challenges

Non-orthogonal multiple access (NOMA) is a promising radio access technology for the 5G wireless systems. The core of NOMA is to support multiple users in the same resource block via power or code domain multiplexing, which provides great enhancement in spectrum efficiency and connectivity. Meanwhile, with the recent advance in self-interference (SI) cancelation techniques, full duplex (FD) wireless communication has become a feasible technology enabling radios to receive and transmit simultaneously. This article aims to investigate the combination of these two emerging technologies. At first, several typical scenarios and protocols are presented to illustrate the application of FD technique in NOMA systems. Then, a novel NOMA system with FD base stations (BSs) based on centralized radio access networks (C-RAN) is proposed. Furthermore, power allocation policies are discussed for the proposed scheme, and simulation results are provided to demonstrate its superiority. Finally, challenges and research opportunities of FD NOMA systems are also identified to stimulate the future research.

cs.IT↗
arXiv2019

Stability of the conical Kähler-Ricci flows on Fano manifolds

In this paper, we study the stability of the conical Kähler-Ricci flows on Fano manifolds. That is, if there exists a conical Kähler-Einstein metric with cone angle $2πβ$ along the divisor, then for any $β'$ sufficiently close to $β$, the corresponding conical Kähler-Ricci flow converges to a conical Kähler-Einstein metric with cone angle $2πβ'$ along the divisor. Here, we only use the condition that the Log Mabuchi energy is bounded from below. This is a weaker condition than the properness that we have adopted to study the convergence before. As corollaries, we give parabolic proofs of Donaldson's openness theorem and his existence conjecture for the conical Kähler-Einstein metrics with positive Ricci curvatures.

math.DG↗
arXiv2019

Hypersurfaces of the homogeneous nearly Kähler $\mathbb{S}^6$ and $\mathbb{S}^3\times\mathbb{S}^3$ with anticommutative structure tensors

Each hypersurface of a nearly Kähler manifold is naturally equipped with two tensor fields of $(1,1)$-type, namely the shape operator $A$ and the induced almost contact structure $ϕ$. In this paper, we show that, in the homogeneous NK $\mathbb{S}^6$ a hypersurface satisfies the condition $Aϕ+ϕA=0$ if and only if it is totally geodesic; moreover, similar as for the non-flat complex space forms, the homogeneous nearly Kähler manifold $\mathbb{S}^3\times\mathbb{S}^3$ does not admit a hypersurface that satisfies the condition $Aϕ+ϕA=0$.

math.DG↗
arXiv2019

Atmospheres on Nonsynchronized Eccentric-tilted Exoplanets II: Thermal Light Curves

Thermal light-curve analysis is a powerful approach to probe the thermal structures of exoplanetary atmospheres, which are greatly influenced by the planetary obliquity and eccentricity. Here we investigate the thermal light curves of eccentric-tilted exoplanets across various radiative timescales, eccentricities, obliquities, and viewing geometries using results of shallow-water simulations presented in Ohno $\&$ Zhang (2019). We also achieve an analytical theory of the thermal light curve that can explain general trends in the light curves of tilted exoplanets. For tilted planets in circular orbits, the orbital phase of the flux peak is largely controlled by either the flux from the hot spot projected onto the orbital plane or the pole heated at the summer solstice, depending on the radiative timescale $τ_{\rm rad}$, planetary day $P_{\rm orb}$, and obliquity $θ$. We find that tilted planets potentially produce the flux peak after the secondary eclipse when obliquity is $θ$ > 90 deg for the hot regime $τ_{\rm rad} 18$ deg for the cool regime $τ_{\rm rad} > P_{\rm rot}$. For tilted planets in eccentric orbits, the shape of the light curve is considerably influenced by the heating at the periapse. The flux peak occurring after the secondary eclipse can be used to distinguish tilted planets from nontilted planets when the periapse takes place before the secondary eclipse. Our results could help to constrain exoplanet obliquities in future observations.

astro-ph.EP↗
arXiv2019

Atmospheres on Nonsynchronized Eccentric-tilted Exoplanets I: Dynamical Regimes

Relatively long-period nonsynchronized planets---such as warm Jupiters---potentially retain the primordial rotation, eccentricity, and obliquity that might encapsulate information on planetary climate and formation processes. To date, there has not been a systematic study on climate patterns on these planets that will significantly influence their observations. Here we investigate the atmospheric dynamics of nonsynchronized, fast-rotating exoplanets across various radiative timescales, eccentricities, and obliquities using a shallow water model. The dynamical pattern can be demarcated into five regimes in terms of radiative timescale $τ_{\rm rad}$ and obliquity $θ$. An atmosphere with $τ_{\rm rad}$ shorter than a planetary day usually exhibits a strong day--night temperature contrast and a day-to-night flow pattern. In the intermediate $τ_{\rm rad}$ regime between a planetary day and a year, the atmosphere is dominated by steady temperature and eastward jet patterns for $θ$ < 18 deg but shows a strong seasonal variation for $θ$ > 18 deg because the polar region undergoes an intense heating at around the summer solstice. If $τ_{\rm rad}$ is longer than a year, seasonal variation is very weak. In this regime, eastward jets are developed for $θ$ < 54 deg and westward jets are developed for $θ$ > 54 deg. These dynamical regimes are also applicable to the planets in eccentric orbits. The large effects of exoplanetary obliquities on circulation patterns might offer observational signatures, which will be investigated in Paper II of this study.

astro-ph.EP↗