• Recently, (Blanchet, Kang, and Murhy 2016, and Blanchet, and Kang 2017) showed that several machine learning algorithms, such as square-root Lasso, Support Vector Machines, and regularized logistic regression, among many others, can be represented exactly as distributionally robust optimization (DRO) problems. The distributional uncertainty is defined as a neighborhood centered at the empirical distribution. We propose a methodology which learns such neighborhood in a natural data-driven way. We show rigorously that our framework encompasses adaptive regularization as a particular case. Moreover, we demonstrate empirically that our proposed methodology is able to improve upon a wide range of popular machine learning estimators.
  • Topological superfluids protected by mirror and time-reversal symmetries are exotic states of matter possessing Majorana Kramers pairs (MKPs), yet their realizations have long been hindered by the requirement of unconventional pairing. We propose to realize such a topological superfluid by utilizing $s$-wave pairing and emergent mirror and time-reversal symmetries in two coupled 1D ultracold atomic Fermi gases with spin-orbit coupling. By stacking such systems into 2D, we discover topological and Dirac-nodal superfluids hosting distinct MKP flat bands. We show that the emergent symmetries make the MKPs and their flat bands stable against pairing fluctuations that otherwise annihilate paired Majoranas. Exploiting new experimental developments, our scheme provides a unique platform for exploring MKPs and their applications in quantum computation.
  • The recent discovery of triply degenerate points (TDPs) in topological materials has opened a new perspective toward the realization of novel quasiparticles without counterparts in quantum field theory. The emergence of such protected nodes is often attributed to spin-vector-momentum couplings. Here we show that the interplay between spin-tensor- and spin-vector-momentum couplings can induce three types of TDPs, classified by different monopole charges ($\mathcal{C}=\pm 2,\pm 1,0$). A Zeeman field can lift them into Weyl points with distinct numbers and charges. Different TDPs of the same type are connected by intriguing Fermi arcs at surfaces, and transitions between different types are accompanied by level crossings along high-symmetry lines. We further propose an experimental scheme to realize such TDPs in cold-atom optical lattices. Our results provide a framework for studying spin-tensor-momentum coupling-induced TDPs and other exotic quasiparticles.
  • For the task of subdecimeter aerial imagery segmentation, the fine-grained semantic segmentation results are usually difficult to obtain because of complex remote sensing contents and optical conditions. In addition, remote sensing imagery has inherent limitations of imbalanced class distribution. Recently, convolutional neural networks (CNNs) have shown outstanding performance on this task. In this paper, we propose the TreeSegNet to solve the class imbalance problem and further improve the accuracy in the metrics' point of view. Based on the infrastructure of DeepUNet, a Tree-CNN model in which each node represents a ResNeXt unit is constructed automatically according to confusion matrix and minimum graph cut algorithm. By transporting feature maps by concatenating connections, the Tree-CNN block fuses the multiscale features and learning the best weights for the model. In the experiments on ISPRS 2D semantic labeling Potsdam dataset, the results gotten by TreeSegNet are better than the opened state-of-the-art methods. The F1 measure scores of classes are improved especially for those classes that are easily confused. Completely and detailed comparison and analysis are performed to show that the improvement is brought by the construction and the embedding of the Tree-CNN module.
  • Van der Waals (vdW) heterostructures synthesized through the chemical vapor deposition (CVD) method allow creation and tuning of intriguing electronic and optical properties of two- dimensional (2D) materials, the knowledge of which is critical for a wide range of potential applications. Here we report our scanning tunneling microscopy/spectroscopy (STM/STS) study of as-grown MoS2 monolayer and WS2/MoS2 heterobilayer on SiO2. The heterobilayer appears smoother than the MoS2 monolayer, with root mean square (RMS) roughness of 0.230 +- 0.021 nm in the former and 0.329 +- 0.033 nm in the latter. For the first time, to our knowledge, we directly observed a continuous interface between the MoS2 monolayer and the top layer of the heterobilayer with atomic resolution. This finding contrasts to the previously reported open edges in the top layer of the heterobilayer. Our STS results and density functional theory (DFT) calculations revealed the band gaps of the heterobilayer and the MoS2 monolayer.
  • The Alternating Direction Method of Multipliers (ADMM) decoding of Low Density Parity Check (LDPC) codes has received many attentions due to its excellent performance at the error floor region. In this paper, we develop a parameter-free decoder based on Linear Program (LP) decoding by replacing the binary constraint with the intersection of a box and an $\ell_p$ sphere. An efficient $\ell_2$-box ADMM is designed to handle this model in a distributed fashion. Numerical experiments demonstrate that our decoder attains better adaptability to different Signal-to-Noise Ratio and channels.
  • Providing backhaul connectivity for macro and pico base stations (BSs) constitutes a significant share of infrastructure costs in future heterogeneous networks (HetNets). To address this issue, the emerging idea of flexible backhaul is proposed. Under this architecture, not all the pico BSs are connected to the backhaul, resulting in a significant reduction in the infrastructure costs. In this regard, pico BSs without backhaul connectivity need to communicate with their nearby BSs in order to have indirect accessibility to the backhaul. This makes the radio resource management (RRM) in such networks more complex and challenging. In this paper, we address the problem of cross-layer RRM in HetNets with flexible backhaul. We formulate this problem as a two-timescale non-convex stochastic optimization which jointly optimizes flow control, routing, interference mitigation and link scheduling in order to maximize a generic network utility. By exploiting a hidden convexity of this non-convex problem, we propose an iterative algorithm which converges to the global optimal solution. The proposed algorithm benefits from low complexity and low signalling, which makes it scalable. Moreover, due to the proposed two-timescale design, it is robust to the backhaul signalling latency as well. Simulation results demonstrate the significant performance gain of the proposed solution over various baselines.
  • In this paper, a simple field function is presented for facilitating the solution of complex and dynamic fluid-solid systems on Cartesian grids with interface-resolved fluid-fluid, fluid-solid, and solid-solid interactions. For a Cartesian-grid-discretized computational domain segmented by a set of solid bodies, this field function explicitly tracks each subdomain with multiple resolved interfacial node layers. As a result, the presented field function enables low-memory-cost multidomain node mapping, efficient node remapping, fast collision detection, and expedient surface force integration. Implementation algorithms for the field function and its described functionalities are also presented. Equipped with a deterministic multibody collision model, numerical experiments involving fluid-solid systems with flow conditions ranging from subsonic to supersonic states are conducted to validate and illustrate the applicability of the proposed field function.
  • In this paper, a novel immersed boundary method is developed, validated, and applied. Through devising a second-order three-step flow reconstruction scheme, the proposed method is able to enforce the Dirichlet, Neumann, Robin, and Cauchy boundary conditions in a straightforward and consistent manner. Equipped with a fluid-solid coupling framework that integrates high-order temporal and spatial discretization schemes, numerical experiments concerning flow involving stationary and moving objects, convex and concave geometries, no-slip and slip wall boundary conditions, as well as subsonic and supersonic motions are conducted to validate the method. It is demonstrated that the proposed method can provide efficient, accurate, and robust boundary treatment for solving flow with arbitrarily irregular and moving geometries on Cartesian grids.
  • The topological valley Hall effect was predicted as a consequence of the bulk topology of electronic systems. Now it has been observed in photonic crystals, showing that both topology and valley are innate to classical as well as quantum systems. The discovery of counterflow in opposite valleys not only demonstrates that the valley can be a novel carrier of information and energy --- particularly valuable for systems without charge and spin --- but also exemplifies that symmetry-protected band topology can be universal to both quantum and classical systems.
  • Smart contracts are applications that execute on blockchains. Today they manage billions of dollars in value and motivate visionary plans for pervasive blockchain deployment. While smart contracts inherit the availability and other security assurances of blockchains, however, they are impeded by blockchains' lack of confidentiality and poor performance. We present Ekiden, a system that addresses these critical gaps by combining blockchains with Trusted Execution Environments (TEEs), such as Intel SGX. Capable of operating on any desired blockchain, Ekiden permits concurrent, off-chain execution of smart contracts within TEE-backed compute nodes, yielding high performance, low cost, and confidentiality for sensitive data. Ekiden enforces a strong set of security and availability properties. By maintaining on-chain state, it achieves consistency, meaning a single authoritative sequence of state transitions, and availability, meaning contracts can survive the failure of compute nodes. Ekiden is anchored in a formal security model expressed as an ideal functionality. We prove the security of the corresponding implemented protocol in the UC framework. Our implementation of Ekiden supports contract development in Rust and the Ethereum Virtual Machine (EVM). We present experiments for applications including machine learning models, poker, and cryptocurrency tokens. Ekiden is designed to support multiple underlying blockchains. When building on top of Tendermint for consensus, we obtain example performance of 600x more throughput and 400x less latency at 1000x less cost than on the Ethereum mainnet. When used with Ethereum as the backing blockchain, Ekiden still costs less than on-chain execution and supports contract confidentiality.
  • Majorana bound states often occur at the end of 1D topological superconductor or at the $\pi$ Josephson junction mediated by a helical edge state. Validated by a new bulk invariant and an intuitive edge argument, we show the emergence of one Majorana Kramers pair at each corner of a square-shaped 2D topological insulator proximitized by an $s_\pm$-wave (e.g., Fe-based) superconductor. We obtain a phase diagram that emphasizes the roles of bulk parameters and edge orientations. We propose several experimental realizations in lattice-matched candidate materials. Our scheme offers a high-temperature platform for exploring higher-order non-Abelian quasiparticles.
  • Deep convolutional neural networks (CNNs) have greatly improved the Face Recognition (FR) performance in recent years. Almost all CNNs in FR are trained on the carefully labeled datasets containing plenty of identities. However, such high-quality datasets are very expensive to collect, which restricts many researchers to achieve state-of-the-art performance. In this paper, we propose a framework, called SeqFace, for learning discriminative face features. Besides a traditional identity training dataset, the designed SeqFace can train CNNs by using an additional dataset which includes a large number of face sequences collected from videos. Moreover, the label smoothing regularization (LSR) and a new proposed discriminative sequence agent (DSA) loss are employed to enhance discrimination power of deep face features via making full use of the sequence data. Our method achieves excellent performance on Labeled Faces in the Wild (LFW), YouTube Faces (YTF), only with a single ResNet. The code and models are publicly available on-line (https://github.com/huangyangyu/SeqFace).
  • Monolayer transition metal dichalcogenides are recently emerged 2D electronic systems with various novel properties, such as spin-valley locking, circular dichroism, valley Hall effects, Ising superconductivity. The reduced dimensionality and large effective masses further produce unconventional many-body interaction effects. Although recent hole transport measurements in WSe$_2$ indicate strong interactions in the valence bands, many-body interaction effects, particularly in the conduction bands, remain elusive to date. Here, for the first time, we perform transport measurements up to a magnetic field of $29$T to study the massive Dirac electron Landau levels (LL) in layer-polarized MoS$_2$ samples with mobilities of $22000$cm$^2$/(V$\cdot$s) at $1.5$K and densities of $\sim10^{12}$cm$^{-2}$. With decreasing the density, we observe LL crossing induced valley ferrimagnet-to-ferromagnet transitions, as a result of the interaction enhancement of the g-factor from $5.64$ to $21.82$. Near integer ratios of Zeeman-to-cyclotron energies, we discover LL anticrossings due to the formation of quantum Hall Ising ferromagnets, the valley polarizations of which appear to be reversible by tuning the density or an in-plane magnetic field. Our results provide compelling evidence for many-body interaction effects in the conduction bands of monolayer MoS$_2$ and establish a fertile ground for exploring strongly correlated phenomena of massive Dirac electrons.
  • In the Dirac-Weyl equation that describes massless neutrino propagation in the minimal Standard Model, the ($2\times 2$ equivalence of the) gamma matrices convert Weyl spinors into spacetime tensors, and vice versa. They can thus be regarded as amalgamations of three different types of mappings, one that connects particle spinors directly forming representations to internal gauge symmetries to their spacetime counterparts that are embodied by null flags, another that translates the spacetime spinors into their corresponding tensors expressed in an orthonormal tetrad, and finally a purely tensorial transformation into the coordinate tetrad. The splitting of spinors into particle and spacetime varieties is not usually practised, but we advocate its adoption for better physical clarity, in terms of distinguishing internal and spacetime transformations, and also for understanding the scattering of neutrinos by spacetime curvature. We construct the basic infrastructure required for this task, and provide a worked example for the Schwarzschild spacetime. Our investigation also uncovers a possible under-determinacy in the flavoured Dirac-Weyl equation, which could serve as a new incision point for introducing flavour oscillation mechanisms.
  • Plasma-filled pulsar magnetospheres contain thin current sheets wherein the charged particles are accelerated by magnetic reconnections to travel at ultra-relativistic speeds. On the other hand, the plasma frequency of the more regular force-free regions of the magnetosphere rests almost precisely on the upper limit of radio frequencies, with the cyclotron frequency being far higher due to the strong magnetic field. This combination produces a peculiar situation, whereby radio-frequency waves can travel at subluminal speeds without becoming evanescent. The conditions are thus conducive to Cerenkov radiation originating from current sheets, which could plausibly serve as a coherent radio emission mechanism. In this paper we aim to provide a portrait of the relevant processes involved, and show that this mechanism can possibly account for some of the most salient features of the observed radio signals.
  • Recently, the topological physics in artificial crystals for classical waves has become an emerging research area. In this Letter, we propose a unique bilayer design of sonic crystals that are constructed by two layers of coupled hexagonal array of triangular scatterers. Assisted by the additional layer degree of freedom, a rich topological phase diagram is achieved by simply rotating scatterers in both layers. Under a unified theoretical framework, two kinds of valley-projected topological acoustic insulators are distinguished analytically, i.e., the layer-mixed and layer-polarized topological valley Hall phases, respectively. The theory is evidently confirmed by our numerical and experimental observations of the nontrivial edge states that propagate along the interfaces separating different topological phases. Various applications such as sound communications in integrated devices, can be anticipated by the intriguing acoustic edge states enriched by the layer information.
  • Recently, an instability due to the nonlinear coupling of p-modes to g-modes in tidally deformed neutron stars in coalescing binaries has been studied in some detail. The result is significant because it could influence the inspiral and leave an imprint on the gravitational wave signal that depends on the neutron star equation of state (EOS). Because of its potential importance, the details of the instability should be further elucidated and its sensitivity to the EOS should be investigated. To this end, we carry out a numerical analysis with six representative EOSs for both static and non-static tides. We confirm that the absence of the p-g instability under static tides, as well as its return under non-static tides, is generic across EOSs, and further reveal a new contribution to it that becomes important for moderately high-order p-g pairs (previous studies concentrated on very high order modes), whose associated coupling strength can vary by factors of ~10-100 depending on the EOS. We find that, for stars with stiffer EOSs and smaller buoyancy frequencies, the instability onsets earlier in the inspiral and the unstable modes grow faster. These results suggest that the instability's impact on the gravitational wave signal might be sensitive to the neutron star EOS. To fully assess this prospect, future studies will need to investigate its saturation as a function of the EOS and the binary parameters.
  • Cd3As2 is a three-dimensional topological Dirac semimetal with connected Fermi-arc surface states. It has been suggested that topological superconductivity can be achieved in the nontrivial surface states of topological materials by utilizing the superconductor proximity effect. Here we report observations of both {\pi} and 4{\pi} periodic supercurrents in aluminum-Cd3As2-aluminum Josephson junctions. The {\pi} period is manifested by both the magnetic-field dependence of the critical supercurrent and the appearance of half-integer Shapiro steps in the a.c. Josephson effect. Our macroscopic theory suggests that the {\pi} period arises from interference between the induced bulk superconductivity and the induced Fermi arc surface superconductivity. The 4{\pi} period is manifested by the missing first Shapiro steps and is expected for topological superconductivity.
  • Finding the electromagnetic (EM) counterpart of binary compact star merger, especially the binary neutron star (BNS) merger, is critically important for gravitational wave (GW) astronomy, cosmology and fundamental physics. On Aug. 17, 2017, Advanced LIGO and \textit{Fermi}/GBM independently triggered the first BNS merger, GW170817, and its high energy EM counterpart, GRB 170817A, respectively, resulting in a global observation campaign covering gamma-ray, X-ray, UV, optical, IR, radio as well as neutrinos. The High Energy X-ray telescope (HE) onboard \textit{Insight}-HXMT (Hard X-ray Modulation Telescope) is the unique high-energy gamma-ray telescope that monitored the entire GW localization area and especially the optical counterpart (SSS17a/AT2017gfo) with very large collection area ($\sim$1000 cm$^2$) and microsecond time resolution in 0.2-5 MeV. In addition, \textit{Insight}-HXMT quickly implemented a Target of Opportunity (ToO) observation to scan the GW localization area for potential X-ray emission from the GW source. Although it did not detect any significant high energy (0.2-5 MeV) radiation from GW170817, its observation helped to confirm the unexpected weak and soft nature of GRB 170817A. Meanwhile, \textit{Insight}-HXMT/HE provides one of the most stringent constraints (~10$^{-7}$ to 10$^{-6}$ erg/cm$^2$/s) for both GRB170817A and any other possible precursor or extended emissions in 0.2-5 MeV, which help us to better understand the properties of EM radiation from this BNS merger. Therefore the observation of \textit{Insight}-HXMT constitutes an important chapter in the full context of multi-wavelength and multi-messenger observation of this historical GW event.
  • The basic concept of multi-dimensional limiting process (MLP) on unstructured grids is inherited and modified for improving shock stabilities and reducing numerical dissipation on smooth regions. A relaxed version of MLP condition, simply named as weak-MLP, is proposed for reducing dissipation. Moreover, a stricter condition, that is the strict-MLP condition, is proposed to enhance the numerical stability. The maximum/minimum principle is fulfilled by both the strict- and weak-MLP condition. A differentiable pressure weight function is applied for the combination of two novel conditions, and thus the modified limiter is named as MLP-pw(pressure-weighted). A series of numerical test cases show that MLP-pw limiter has improved stability and convergence, especially in hypersonic simulations. Furthermore, the limiter also shows lower dissipation in regions without significant pressure transition. Therefore, MLP-pw limiter can capture contact discontinuity and expansion accurately.
  • Atomically thin transition metal dichalcogenides (TMDCs) such as MoS$_2$ and WSe$_2$ are emerging as a new platform for exploring many-body effects. Coulomb interactions are markedly enhanced in these materials because of the reduced screening and the large Wigner-Seitz radii. Although many-body excitonic effects in TMDCs have been extensively studied by optical means, not until recently did probing their strongly correlated electronic effects become possible in transport. Here, in p-type few-layer WSe$_2$ we observe highly density-dependent quantum Hall states of {\Gamma} valley holes below 12 T, whose predominant sequences alternate between odd- and even-integers. By tilting the magnetic field to induce Landau level crossings, we show that the strong Coulomb interaction enhances the Zeeman-to-cyclotron energy ratio from 2.67 to 3.55 as the density is reduced from 5.7 to 4.0$\times$10$^{12}$ cm$^{-2}$, giving rise to the even-odd alternation. Unprecedentedly, this indicates a 4.8 times enhancement of the g-factor over its band theory value at a density as high as 4.0$\times$10$^{12}$ cm$^{-2}$. Our findings unambiguously demonstrate that p-type few-layer WSe$_2$ is a superior platform for exploring strongly correlated electronic phenomena, opening a new perspective for realizing the elusive Wigner crystallization at a moderate density.
  • Scalar, vector and tensor harmonics on the three-sphere were introduced originally to facilitate the study of various problems in gravitational physics. These harmonics are defined as eigenfunctions of the covariant Laplace operator which satisfy certain divergence and trace identities, and ortho-normality conditions. This paper provides a summary of these properties, along with a new notation that simplifies and clarifies some of the key expressions. Practical methods are described for accurately and efficiently computing these harmonics numerically, and test results are given that illustrate how well the analytical identities are satisfied by the harmonics computed numerically in this way.
  • Semantic segmentation is a fundamental research in remote sensing image processing. Because of the complex maritime environment, the sea-land segmentation is a challenging task. Although the neural network has achieved excellent performance in semantic segmentation in the last years, there are a few of works using CNN for sea-land segmentation and the results could be further improved. This paper proposes a novel deep convolution neural network named DeepUNet. Like the U-Net, its structure has a contracting path and an expansive path to get high resolution output. But differently, the DeepUNet uses DownBlocks instead of convolution layers in the contracting path and uses UpBlock in the expansive path. The two novel blocks bring two new connections that are U-connection and Plus connection. They are promoted to get more precise segmentation results. To verify our network architecture, we made a new challenging sea-land dataset and compare the DeepUNet on it with the SegNet and the U-Net. Experimental results show that DeepUNet achieved good performance compared with other architectures, especially in high-resolution remote sensing imagery.
  • The long-thought charge Kondo effects have recently been experimentally realized in the quantum Hall regime. This experiment, supported by numerics, exemplifies the realization of two-channel Kondo state, a non-Fermi Liquid, and its crossover to the one-channel counterpart, a Fermi liquid. Scaling up such a platform, we find a hierarchy of non-Fermi Liquids and their tunable crossovers based on a renormalization group analysis. Utilizing results from a conformal field theory, we further examine the universal conductances of this strongly correlated system and their finite temperature scaling, which elucidate the sharp distinctions between charge and spin Kondo physics.