arXiv subjects

Cheng Chen

Publications and source records attributed to Cheng Chen.

At least 361 records · Page 20Linked to original sources

arXiv2020

Large-Scale Optimal Transport via Adversarial Training with Cycle-Consistency

Recent advances in large-scale optimal transport have greatly extended its application scenarios in machine learning. However, existing methods either not explicitly learn the transport map or do not support general cost function. In this paper, we propose an end-to-end approach for large-scale optimal transport, which directly solves the transport map and is compatible with general cost function. It models the transport map via stochastic neural networks and enforces the constraint on the marginal distributions via adversarial training. The proposed framework can be further extended towards learning Monge map or optimal bijection via adopting cycle-consistency constraint(s). We verify the effectiveness of the proposed method and demonstrate its superior performance against existing methods with large-scale real-world applications, including domain adaptation, image-to-image translation, and color transfer.

cs.CV↗

arXiv2020

Atomic line defects and zero-energy end states in monolayer Fe(Te,Se) high-temperature superconductors

Majorana zero-energy bound states (ZEBSs) have been proposed to exist at the ends of one-dimensional Rashba nanowires proximity-coupled to an s-wave superconductor in an external magnetic field induced Zeeman field. Such hybrid structures have been a central platform in the search for non-Abelian Majorana zero modes (MZMs) toward fault-tolerant topological quantum computing. Here we report the discovery of ZEBSs simultaneously appearing at each end of a one-dimensional atomic line defect in monolayer iron-based high-temperature superconductor FeTe0.5Se0.5 films grown on SrTiO3(001) substrates. The spectroscopic properties of the ZEBSs, including the temperature and tunneling barrier dependences, as well as their fusion induced by coupling on line defects of different lengths are found to be robust and consistent with those of the MZMs. These observations suggest a realization of topological Shockley defects at the ends of an atomic line defect in a two-dimensional s-wave superconductor that can host a Kramers pair of MZMs protected by time-reversal symmetry along the chain. Our findings reveal an unprecedented class of topological line defect excitations in two-dimensional superconductor FeTe0.5Se0.5 monolayer films and offer an advantageous platform for generating topological zero-energy excitations at higher operating temperatures, in a single material, and under zero external magnetic field.

cond-mat.supr-con↗

arXiv2020

Zero-energy bound states in the high-temperature superconductors at the two-dimensional limit

Majorana zero modes (MZMs) that obey the non-Abelian statistics have been intensively investigated for potential applications in topological quantum computing. The prevailing signals in tunneling experiments "fingerprinting" the existence of MZMs are the zero-energy bound states (ZEBSs). However, nearly all of the previously reported ZEBSs showing signatures of the MZMs are observed in difficult-to-fabricate heterostructures at very low temperatures and additionally require applied magnetic field. Here, by using in-situ scanning tunneling spectroscopy, we detect the ZEBSs upon the interstitial Fe adatoms deposited on two different high-temperature superconducting one-unit-cell-thick iron chalcogenides on SrTiO3(001). The spectroscopic results resemble the phenomenological characteristics of the MZMs inside the vortex cores of topological superconductors. Our experimental findings may extend the MZM explorations in connate topological superconductors towards an applicable temperature regime and down to the two-dimensional limit. While a concrete understanding of the observations is lacking, possible explanations involving novel 2D superconducting states with spin-orbit coupling, spontaneous nucleation of anomalous vortices at the magnetic sites, and noncoplanar magnetic ordering may further stimulate theoretical understandings of the scarcely captured ZEBSs in strongly correlated systems with multiband Cooper pairing.

cond-mat.supr-con↗

arXiv2020

Formation of the polar debris disc around 99 Herculis

We investigate the formation mechanism for the observed nearly polar aligned (perpendicular to the binary orbital plane) debris ring around the eccentric orbit binary 99 Herculis. An initially inclined nonpolar debris ring or disc will not remain flat and will not evolve to a polar configuration, due to the effects of differential nodal precession that alter its flat structure. However, a gas disc with embedded well coupled solids around the eccentric binary may evolve to a polar configuration as a result of pressure forces that maintain the disc flatness and as a result of viscous dissipation that allows the disc to increase its tilt. Once the gas disc disperses, the debris disc is in a polar aligned state in which there is little precession. We use three-dimensional hydrodynamical simulations, linear theory, and particle dynamics to study the evolution of a misaligned circumbinary gas disc and explore the effects of the initial disc tilt, mass, and size. We find that for a wide range of parameter space, the polar alignment timescale is shorter than the lifetime of the gas disc. Using the observed level of alignment of 3 deg. from polar, we place an upper limit on the mass of the gas disc of about 0.014 M_sun at the time of gas dispersal. We conclude that the polar debris disc around 99 Her can be explained as the result of an initially moderately inclined gas disc with embedded solids. Such a disc may provide an environment for the formation of polar planets.

astro-ph.EP↗

arXiv2020

Robust Multimodal Brain Tumor Segmentation via Feature Disentanglement and Gated Fusion

Accurate medical image segmentation commonly requires effective learning of the complementary information from multimodal data. However, in clinical practice, we often encounter the problem of missing imaging modalities. We tackle this challenge and propose a novel multimodal segmentation framework which is robust to the absence of imaging modalities. Our network uses feature disentanglement to decompose the input modalities into the modality-specific appearance code, which uniquely sticks to each modality, and the modality-invariant content code, which absorbs multimodal information for the segmentation task. With enhanced modality-invariance, the disentangled content code from each modality is fused into a shared representation which gains robustness to missing data. The fusion is achieved via a learning-based strategy to gate the contribution of different modalities at different locations. We validate our method on the important yet challenging multimodal brain tumor segmentation task with the BRATS challenge dataset. With competitive performance to the state-of-the-art approaches for full modality, our method achieves outstanding robustness under various missing modality(ies) situations, significantly exceeding the state-of-the-art method by over 16% in average for Dice on whole tumor segmentation.

cs.CV↗

arXiv2020

Asteroid belt survival through stellar evolution: dependence on the stellar mass

Polluted white dwarfs are generally accreting terrestrial-like material that may originate from a debris belt like the asteroid belt in the solar system. The fraction of white dwarfs that are polluted drops off significantly for white dwarfs with masses $M_{\rm WD}\gtrsim 0.8\,\rm M_\odot$. This implies that asteroid belts and planetary systems around main-sequence stars with mass $M_{\rm MS}\gtrsim 3\,\rm M_\odot$ may not form because of the intense radiation from the star. This is in agreement with current debris disc and exoplanet observations. The fraction of white dwarfs that show pollution also drops off significantly for low mass white dwarfs $(M_{\rm WD}\lesssim 0.55\,\rm M_\odot)$. However, the low-mass white dwarfs that do show pollution are not currently accreting but have accreted in the past. We suggest that asteroid belts around main sequence stars with masses $M_{\rm MS}\lesssim 2\,\rm M_\odot$ are not likely to survive the stellar evolution process. The destruction likely occurs during the AGB phase and could be the result of interactions of the asteroids with the stellar wind, the high radiation or, for the lowest mass stars that have an unusually close-in asteroid belt, scattering during the tidal orbital decay of the inner planetary system.

astro-ph.EP↗

arXiv2020

Unsupervised Bidirectional Cross-Modality Adaptation via Deeply Synergistic Image and Feature Alignment for Medical Image Segmentation

Unsupervised domain adaptation has increasingly gained interest in medical image computing, aiming to tackle the performance degradation of deep neural networks when being deployed to unseen data with heterogeneous characteristics. In this work, we present a novel unsupervised domain adaptation framework, named as Synergistic Image and Feature Alignment (SIFA), to effectively adapt a segmentation network to an unlabeled target domain. Our proposed SIFA conducts synergistic alignment of domains from both image and feature perspectives. In particular, we simultaneously transform the appearance of images across domains and enhance domain-invariance of the extracted features by leveraging adversarial learning in multiple aspects and with a deeply supervised mechanism. The feature encoder is shared between both adaptive perspectives to leverage their mutual benefits via end-to-end learning. We have extensively evaluated our method with cardiac substructure segmentation and abdominal multi-organ segmentation for bidirectional cross-modality adaptation between MRI and CT images. Experimental results on two different tasks demonstrate that our SIFA method is effective in improving segmentation performance on unlabeled target images, and outperforms the state-of-the-art domain adaptation approaches by a large margin.

eess.IV↗

arXiv2019

Archimedean Non-vanishing, Cohomological Test Vectors, and Standard $L$-functions of $\mathrm{GL}_{2n}$: Real Case

The standard $L$-functions of $\mathrm{GL}_{2n}$ expressed in terms of the Friedberg-Jacquet global zeta integrals have better structure for arithmetic applications, due to the relation of the linear periods with the modular symbols. The most technical obstacles towards such arithmetic applications are (1) non-vanishing of modular symbols at infinity and (2) the existance or construction of uniform cohomological test vectors. Problem (1) is also called the non-vanishing hypothesis at infinity, which was proved by Binyong Sun, by establishing the existence of certain cohomological test vectors. In this paper, we explicitly construct an archimedean local integral that produces a new type of a twisted linear functional $Λ_{s,χ}$, which, when evaluated with our explicitly constructed cohomological vector, is equal to the local twisted standard $L$-function $L(s,π\otimesχ)$ as a meromorphic function of $s\in \mathbb{C}$. With the relations between linear models and Shalika models, we establish (1) with an explicitly constructed cohomological vector, and hence recovers a non-vanishing result of Binyong Sun via a completely different method. Our main result indicates a complete solution to (2), which will be presented in a paper of Dihua Jiang, Binyong Sun and Fangyang Tian with full details and with applications to the global period relations for the twisted standard $L$-functions at critical places.

math.RT↗

arXiv2019

Multi-Instance Multi-Scale CNN for Medical Image Classification

Deep learning for medical image classification faces three major challenges: 1) the number of annotated medical images for training are usually small; 2) regions of interest (ROIs) are relatively small with unclear boundaries in the whole medical images, and may appear in arbitrary positions across the x,y (and also z in 3D images) dimensions. However often only labels of the whole images are annotated, and localized ROIs are unavailable; and 3) ROIs in medical images often appear in varying sizes (scales). We approach these three challenges with a Multi-Instance Multi-Scale (MIMS) CNN: 1) We propose a multi-scale convolutional layer, which extracts patterns of different receptive fields with a shared set of convolutional kernels, so that scale-invariant patterns are captured by this compact set of kernels. As this layer contains only a small number of parameters, training on small datasets becomes feasible; 2) We propose a "top-k pooling" to aggregate the feature maps in varying scales from multiple spatial dimensions, allowing the model to be trained using weak annotations within the multiple instance learning (MIL) framework. Our method is shown to perform well on three classification tasks involving two 3D and two 2D medical image datasets.

cs.CV↗

arXiv2019

A Stochastic Proximal Point Algorithm for Saddle-Point Problems

We consider saddle point problems which objective functions are the average of $n$ strongly convex-concave individual components. Recently, researchers exploit variance reduction methods to solve such problems and achieve linear-convergence guarantees. However, these methods have a slow convergence when the condition number of the problem is very large. In this paper, we propose a stochastic proximal point algorithm, which accelerates the variance reduction method SAGA for saddle point problems. Compared with the catalyst framework, our algorithm reduces a logarithmic term of condition number for the iteration complexity. We adopt our algorithm to policy evaluation and the empirical results show that our method is much more efficient than state-of-the-art methods.

cs.LG↗

arXiv2019

Orbital dynamics of circumbinary planets

We investigate the dynamics of a nonzero mass, circular orbit planet around an eccentric orbit binary for various values of the binary eccentricity, binary mass fraction, planet mass, and planet semi--major axis by means of numerical simulations. Previous studies investigated the secular dynamics mainly by approximate analytic methods. In the stationary inclination state, the planet and binary precess together with no change in relative tilt. For both prograde and retrograde planetary orbits, we explore the conditions for planetary orbital libration versus circulation and the conditions for stationary inclination. As was predicted by analytic models, for sufficiently high initial inclination, a prograde planet's orbit librates about the stationary tilted state. For a fixed binary eccentricity, the stationary angle is a monotonically decreasing function of the ratio of the planet--to--binary angular momentum $j$. The larger $j$, the stronger the evolutionary changes in the binary eccentricity and inclination. We also calculate the critical tilt angle that separates the circulating from the librating orbits for both prograde and retrograde planet orbits. The properties of the librating orbits and stationary angles are quite different for prograde versus retrograde orbits. The results of the numerical simulations are in very good quantitative agreement with the analytic models. Our results have implications for circumbinary planet formation and evolution.

astro-ph.EP↗

arXiv2019

Millemeter-Wave Fixed Wireless Access Using IEEE 802.11ay

IEEE 802.11ay defines new PHY and MAC specifications that enable 100 Gbps communications in the 60 GHz millimeter-wave (mmWave) band. Among the various use cases supported by IEEE 802.11ay, fixed wireless access, a cost-efficient high-performance alternative and/or complement to conventional fixed access, differentiates itself due to its unique requirements and characteristics. In this article, our goal is to identify and describe key elements incorporated into IEEE 802.11ay, including scheduling, beamforming, and link maintenance, that efficiently support fixed wireless access. IEEE 802.11ay is thus a viable and strong candidate to form the basis of future generations of standards-compliant (i.e., non-proprietary) mmWave fixed wireless access networks.

cs.NI↗

arXiv2019

Synergistic Image and Feature Adaptation: Towards Cross-Modality Domain Adaptation for Medical Image Segmentation

This paper presents a novel unsupervised domain adaptation framework, called Synergistic Image and Feature Adaptation (SIFA), to effectively tackle the problem of domain shift. Domain adaptation has become an important and hot topic in recent studies on deep learning, aiming to recover performance degradation when applying the neural networks to new testing domains. Our proposed SIFA is an elegant learning diagram which presents synergistic fusion of adaptations from both image and feature perspectives. In particular, we simultaneously transform the appearance of images across domains and enhance domain-invariance of the extracted features towards the segmentation task. The feature encoder layers are shared by both perspectives to grasp their mutual benefits during the end-to-end learning procedure. Without using any annotation from the target domain, the learning of our unified model is guided by adversarial losses, with multiple discriminators employed from various aspects. We have extensively validated our method with a challenging application of cross-modality medical image segmentation of cardiac structures. Experimental results demonstrate that our SIFA model recovers the degraded performance from 17.2% to 73.0%, and outperforms the state-of-the-art methods by a significant margin.

cs.CV↗

arXiv2019

Physical Layer Performance Evaluation of Wireless Infrared-based LiFi Uplink

LiFi (light-fidelity) is recognised as a promising technology for the next generation wireless access network. However, little research effort has been spent on the uplink transmission system in LiFi networks. In this paper, an analytical framework for the performance analysis of an wireless infrared-based LiFi uplink at system level is presented. This work focuses on the performance of a single user who is randomly located in a network. Many important factors in practice are taken into account, such as front-end characteristics, channel path loss, link blockage and user device random orientations. In particular, accurate analytical expressions for the statistics of optical channel path loss has been derived. Based on this path loss statistics, the distribution of signal-to-noise ratio (SNR) and the average achievable data rate are evaluated. A channel factor is defined as a quantity which only depends on the used devices, optical channel and noise power spectral density function (PSD). The result shows that with commercial front-end elements, an average data rate of up to 250 Mbps is achievable.

cs.IT↗

arXiv2019

New-Generation Design-Technology Co-Optimization (DTCO): Machine-Learning Assisted Modeling Framework

In this paper, we propose a machine-learning assisted modeling framework in design-technology co-optimization (DTCO) flow. Neural network (NN) based surrogate model is used as an alternative of compact model of new devices without prior knowledge of device physics to predict device and circuit electrical characteristics. This modeling framework is demonstrated and verified in FinFET with high predicted accuracy in device and circuit level. Details about the data handling and prediction results are discussed. Moreover, same framework is applied to new mechanism device tunnel FET (TFET) to predict device and circuit characteristics. This work provides new modeling method for DTCO flow.

cs.OH↗

arXiv2019

Design of a general scientific CCD simulation and test system

In this paper, we introduce a general scientific CCD simulation and test system which can meet the test requirements of different types of CCD chips from E2V company, such as measuring different signals output from modules of a CCD controller including power supply, fan, temperature control module, crystal oscillator, shutter and clock-bias generator. Furthermore, the video signal of the CCD detector can be simulated and superimposed with random noise to verify the performance of the video sampling circuit of the CCD controller. The simulation and test system was successful used for the CCD controller which was designed for E2V CCD47-20 detector.

physics.ins-det↗

arXiv2019

Robust Frequent Directions with Application in Online Learning

The frequent directions (FD) technique is a deterministic approach for online sketching that has many applications in machine learning. The conventional FD is a heuristic procedure that often outputs rank deficient matrices. To overcome the rank deficiency problem, we propose a new sketching strategy called robust frequent directions (RFD) by introducing a regularization term. RFD can be derived from an optimization problem. It updates the sketch matrix and the regularization term adaptively and jointly. RFD reduces the approximation error of FD without increasing the computational cost. We also apply RFD to online learning and propose an effective hyperparameter-free online Newton algorithm. We derive a regret bound for our online Newton algorithm based on RFD, which guarantees the robustness of the algorithm. The experimental studies demonstrate that the proposed method outperforms state-of-the-art second order online learning algorithms.

cs.LG↗

arXiv2018

PnP-AdaNet: Plug-and-Play Adversarial Domain Adaptation Network with a Benchmark at Cross-modality Cardiac Segmentation

Deep convolutional networks have demonstrated the state-of-the-art performance on various medical image computing tasks. Leveraging images from different modalities for the same analysis task holds clinical benefits. However, the generalization capability of deep models on test data with different distributions remain as a major challenge. In this paper, we propose the PnPAdaNet (plug-and-play adversarial domain adaptation network) for adapting segmentation networks between different modalities of medical images, e.g., MRI and CT. We propose to tackle the significant domain shift by aligning the feature spaces of source and target domains in an unsupervised manner. Specifically, a domain adaptation module flexibly replaces the early encoder layers of the source network, and the higher layers are shared between domains. With adversarial learning, we build two discriminators whose inputs are respectively multi-level features and predicted segmentation masks. We have validated our domain adaptation method on cardiac structure segmentation in unpaired MRI and CT. The experimental results with comprehensive ablation studies demonstrate the excellent efficacy of our proposed PnP-AdaNet. Moreover, we introduce a novel benchmark on the cardiac dataset for the task of unsupervised cross-modality domain adaptation. We will make our code and database publicly available, aiming to promote future studies on this challenging yet important research topic in medical imaging.

cs.CV↗