arXiv subjects

Rui Zhu

Publications and source records attributed to Rui Zhu.

At least 163 records · Page 9Linked to original sources

arXiv2019

Epoxy-inspired Nonlinear Interface Integrating Monolayer Transition-Metal Dichalcogenides with Linear Plasmonic Nanosieves

Although TMDC monolayers offer giant optical nonlinearity within few-angstrom thickness, it is still elusive to modulate and engineer the wavefront of nonlinear emissions. The grain size of high-quality monolayers also restricts possibilities of imparting inhomogeneous or gradient profiles of phases and amplitudes by classical light sources. Plasmonic nanosieves can support huge field enhancement and precise nonlinear phase control with hundred-nanometer pixel-level resolution, however they suffer from intrinsically weak nonlinear susceptibility. In nature, epoxy represents a strong glue whose magic adhesion comes from the bonding of two intrinsic loose and nonsticky resins. Inspired by the nature of epoxy, we report a multifunctional and powerful nonlinear interfaces via integrating transition-metal dichalcogenide (TMDC) monolayers with linear plasmonic nanosieves, which cannot be accessed by either constituents. We experimentally demonstrate orbital angular momentum (OAM) generation, beam steering, versatile polarization control and holograms and of SH emission, with the proposed nonlinear interfaces. In addition, an effective second-harmonic (SH) nonlinearity \c{hi}(2) of ~25 nm/V is obtained for such systems. This designer platform synergizes the complementary advantages of TMDC monolayer, plasmonic nanosieves, geometric phases, and field enhancement, paving a distinct avenue toward designer, multi-functional, and ultra-compact nonlinear optical devices.

physics.optics↗

arXiv2019

Optomechanical Gigahertz Oscillator made of a Two Photon Absorption free piezoelectric III/V semiconductor

Oscillators in the GHz frequency range are key building blocks for telecommunication, timing and positioning applications. Operating directly in the GHz and compactness while keeping high frequency stability, is still an up-to-date challenge. Recently, optomechanical crystals, compact by nature, have demonstrated GHz frequency modes, thus gathering prerequisite features for using them as oscillators. Here we report on the demonstration, in ambient atmospheric conditions, of an optomechanical oscillator designed with an original concept based on bichromatic one-dimensional optomechanical crystal. Self sustained oscillations directly at 3 GHz are routinely achieved with a low optical power threshold of 40 $μW$ and short-term linewidth narrowed down to 100 Hz in agreement with phase noise measurements (-113 dBc/Hz at 1 MHz from the carrier) for free running optomechanical oscillators. This oscillator is made of InGaP, low loss and TPA-free piezoelectric material which makes it valuable for optomechanics.

physics.optics↗

arXiv2019

ScratchDet: Training Single-Shot Object Detectors from Scratch

Current state-of-the-art object objectors are fine-tuned from the off-the-shelf networks pretrained on large-scale classification dataset ImageNet, which incurs some additional problems: 1) The classification and detection have different degrees of sensitivity to translation, resulting in the learning objective bias; 2) The architecture is limited by the classification network, leading to the inconvenience of modification. To cope with these problems, training detectors from scratch is a feasible solution. However, the detectors trained from scratch generally perform worse than the pretrained ones, even suffer from the convergence issue in training. In this paper, we explore to train object detectors from scratch robustly. By analysing the previous work on optimization landscape, we find that one of the overlooked points in current trained-from-scratch detector is the BatchNorm. Resorting to the stable and predictable gradient brought by BatchNorm, detectors can be trained from scratch stably while keeping the favourable performance independent to the network architecture. Taking this advantage, we are able to explore various types of networks for object detection, without suffering from the poor convergence. By extensive experiments and analyses on downsampling factor, we propose the Root-ResNet backbone network, which makes full use of the information from original images. Our ScratchDet achieves the state-of-the-art accuracy on PASCAL VOC 2007, 2012 and MS COCO among all the train-from-scratch detectors and even performs better than several one-stage pretrained methods. Codes will be made publicly available at https://github.com/KimSoybean/ScratchDet.

cs.CV↗

arXiv2019

Constrained Mutual Convex Cone Method for Image Set Based Recognition

In this paper, we propose a method for image-set classification based on convex cone models. Image set classification aims to classify a set of images, which were usually obtained from video frames or multi-view cameras, into a target object. To accurately and stably classify a set, it is essential to represent structural information of the set accurately. There are various representative image features, such as histogram based features, HLAC, and Convolutional Neural Network (CNN) features. We should note that most of them have non-negativity and thus can be effectively represented by a convex cone. This leads us to introduce the convex cone representation to image-set classification. To establish a convex cone based framework, we mathematically define multiple angles between two convex cones, and then define the geometric similarity between the cones using the angles. Moreover, to enhance the framework, we introduce a discriminant space that maximizes the between-class variance (gaps) and minimizes the within-class variance of the projected convex cones onto the discriminant space, similar to the Fisher discriminant analysis. Finally, the classification is performed based on the similarity between projected convex cones. The effectiveness of the proposed method is demonstrated experimentally by using five databases: CMU PIE dataset, ETH-80, CMU Motion of Body dataset, Youtube Celebrity dataset, and a private database of multi-view hand shapes.

cs.CV↗

arXiv2019

Improved Selective Refinement Network for Face Detection

As a long-standing problem in computer vision, face detection has attracted much attention in recent decades for its practical applications. With the availability of face detection benchmark WIDER FACE dataset, much of the progresses have been made by various algorithms in recent years. Among them, the Selective Refinement Network (SRN) face detector introduces the two-step classification and regression operations selectively into an anchor-based face detector to reduce false positives and improve location accuracy simultaneously. Moreover, it designs a receptive field enhancement block to provide more diverse receptive field. In this report, to further improve the performance of SRN, we exploit some existing techniques via extensive experiments, including new data augmentation strategy, improved backbone network, MS COCO pretraining, decoupled classification module, segmentation branch and Squeeze-and-Excitation block. Some of these techniques bring performance improvements, while few of them do not well adapt to our baseline. As a consequence, we present an improved SRN face detector by combining these useful techniques together and obtain the best performance on widely used face detection benchmark WIDER FACE dataset.

cs.CV↗

arXiv2018

SFA: Small Faces Attention Face Detector

In recent year, tremendous strides have been made in face detection thanks to deep learning. However, most published face detectors deteriorate dramatically as the faces become smaller. In this paper, we present the Small Faces Attention (SFA) face detector to better detect faces with small scale. First, we propose a new scale-invariant face detection architecture which pays more attention to small faces, including 4-branch detection architecture and small faces sensitive anchor design. Second, feature maps fusion strategy is applied in SFA by partially combining high-level features into low-level features to further improve the ability of finding hard faces. Third, we use multi-scale training and testing strategy to enhance face detection performance in practice. Comprehensive experiments show that SFA significantly improves face detection performance, especially on small faces. Our real-time SFA face detector can run at 5 FPS on a single GPU as well as maintain high performance. Besides, our final SFA face detector achieves state-of-the-art detection performance on challenging face detection benchmarks, including WIDER FACE and FDDB datasets, with competitive runtime speed. Both our code and models will be available to the research community.

cs.CV↗

arXiv2018

Android Malware Detection using Large-scale Network Representation Learning

With the growth of mobile devices and applications, the number of malicious software, or malware, is rapidly increasing in recent years, which calls for the development of advanced and effective malware detection approaches. Traditional methods such as signature-based ones cannot defend users from an increasing number of new types of malware or rapid malware behavior changes. In this paper, we propose a new Android malware detection approach based on deep learning and static analysis. Instead of using Application Programming Interfaces (APIs) only, we further analyze the source code of Android applications and create their higher-level graphical semantics, which makes it harder for attackers to evade detection. In particular, we use a call graph from method invocations in an Android application to represent the application, and further analyze method attributes to form a structured Program Representation Graph (PRG) with node attributes. Then, we use a graph convolutional network (GCN) to yield a graph representation of the application by embedding the entire graph into a dense vector, and classify whether it is a malware or not. To efficiently train such a graph convolutional network, we propose a batch training scheme that allows multiple heterogeneous graphs to be input as a batch. To the best of our knowledge, this is the first work to use graph representation learning for malware detection. We conduct extensive experiments from real-world sample collections and demonstrate that our developed system outperforms multiple other existing malware detection techniques.

cs.CR↗

arXiv2018

Quantum pumping with adiabatically modulated barriers in three-band pseudospin-1 Dirac-Weyl systems

In this work, pumped currents of the adiabatically-driven double-barrier structure based on the pseudospin-1 Dirac-Weyl fermions are studied. As a result of the three-band dispersion and hence the unique properties of pseudospin-1 Dirac-Weyl quasiparticles, sharp current-direction reversal is found at certain parameter settings especially at the Dirac point of the band structure, where apexes of the two cones touches at the flat band. Such a behavior can be interpreted consistently by the Berry phase of the scattering matrix and the classical turnstile mechanism.

cond-mat.mes-hall↗

arXiv2018

ApolloCar3D: A Large 3D Car Instance Understanding Benchmark for Autonomous Driving

Autonomous driving has attracted remarkable attention from both industry and academia. An important task is to estimate 3D properties(e.g.translation, rotation and shape) of a moving or parked vehicle on the road. This task, while critical, is still under-researched in the computer vision community - partially owing to the lack of large scale and fully-annotated 3D car database suitable for autonomous driving research. In this paper, we contribute the first large-scale database suitable for 3D car instance understanding - ApolloCar3D. The dataset contains 5,277 driving images and over 60K car instances, where each car is fitted with an industry-grade 3D CAD model with absolute model size and semantically labelled keypoints. This dataset is above 20 times larger than PASCAL3D+ and KITTI, the current state-of-the-art. To enable efficient labelling in 3D, we build a pipeline by considering 2D-3D keypoint correspondences for a single instance and 3D relationship among multiple instances. Equipped with such dataset, we build various baseline algorithms with the state-of-the-art deep convolutional neural networks. Specifically, we first segment each car with a pre-trained Mask R-CNN, and then regress towards its 3D pose and shape based on a deformable 3D car model with or without using semantic keypoints. We show that using keypoints significantly improves fitting performance. Finally, we develop a new 3D metric jointly considering 3D pose and 3D shape, allowing for comprehensive evaluation and ablation study. By comparing with human performance we suggest several future directions for further improvements.

cs.CV↗

arXiv2018

Neuromorphic MoS2 memtransistors fabricated by localised helium ion beam irradiation

Two-dimensional layered semiconductors have recently emerged as attractive building blocks for next-generation low-power non-volatile memories. However, challenges remain in the controllable sub-micron fabrication of bipolar resistively switching circuit components from these novel materials. Here we report on the scalable experimental realisation of lateral on-dielectric memtransistors from monolayer single-crystal molybdenum disulfide (MoS2) utilising a focused helium ion beam. Site-specific irradiation with the probe of a helium ion microscope (HIM) allows for the creation of charged defects in the MoS2 lattice. The reversible drift of these locally seeded defects in the applied electric field modulates the resistance of the semiconducting channel, enabling versatile memristive functionality on the nanoscale. We find the device can reliably retain its resistance ratios and set biases for hundreds of switching cycles at sweep frequencies of up to 2.9 V/s with relatively low drain-source biases. We also demonstrate long-term potentiation and depression with sharp habituation that promises application in future neuromorphic architectures. This work advances the down-scaling progress of memristive devices without sacrificing key performance parameters such as power consumption or its applicability for synaptic emulation.

cond-mat.mtrl-sci↗

arXiv2018

A Model Parallel Proximal Stochastic Gradient Algorithm for Partially Asynchronous Systems

Large models are prevalent in modern machine learning scenarios, including deep learning, recommender systems, etc., which can have millions or even billions of parameters. Parallel algorithms have become an essential solution technique to many large-scale machine learning jobs. In this paper, we propose a model parallel proximal stochastic gradient algorithm, AsyB-ProxSGD, to deal with large models using model parallel blockwise updates while in the meantime handling a large amount of training data using proximal stochastic gradient descent (ProxSGD). In our algorithm, worker nodes communicate with the parameter servers asynchronously, and each worker performs proximal stochastic gradient for only one block of model parameters during each iteration. Our proposed algorithm generalizes ProxSGD to the asynchronous and model parallel setting. We prove that AsyB-ProxSGD achieves a convergence rate of $O(1/\sqrt{K})$ to stationary points for nonconvex problems under \emph{constant} minibatch sizes, where $K$ is the total number of block updates. This rate matches the best-known rates of convergence for a wide range of gradient-like algorithms. Furthermore, we show that when the number of workers is bounded by $O(K^{1/4})$, we can expect AsyB-ProxSGD to achieve linear speedup as the number of workers increases. We implement the proposed algorithm on MXNet and demonstrate its convergence behavior and near-linear speedup on a real-world dataset involving both a large model size and large amounts of data.

cs.LG↗

arXiv2018

Statistical Inference Attack Against PHY-layer Key Extraction and Countermeasures

The formal theoretical analysis on channel correlations in both real indoor and outdoor environments are provided in this paper. Moreover, this paper studies empirical statistical inference attacks (SIA) against LSB key extraction, whereby an adversary infers the signature of a target link. Consequently, the secret key extracted from that signature has been recovered by observing the surrounding links. Prior work assumes theoretical link-correlation models for the inference, in contrast, our study does not make any assumption on link correlation. Instead, we take machine learning (ML) methods for link inference based on empirically measured link signatures. ML algorithms have been developed to launch SIAs under various realistic scenarios. Our experimental results have shown that the proposed inference algorithms are still quite effective even without making assumptions on link correlation. In addition, our inference algorithms can reduce the key search space by many orders of magnitudes compared to brute force search. We further propose a countermeasure against the statistical inference attacks, FBCH (forward-backward cooperative key extraction protocol with helpers). In the FBCH, helpers (other trusted wireless nodes) are introduced to provide more randomness in the key extraction. Our experiment results verify the effectiveness of the proposed protocol.

cs.CR↗

arXiv2018

Asynchronous Stochastic Proximal Methods for Nonconvex Nonsmooth Optimization

We study stochastic algorithms for solving nonconvex optimization problems with a convex yet possibly nonsmooth regularizer, which find wide applications in many practical machine learning applications. However, compared to asynchronous parallel stochastic gradient descent (AsynSGD), an algorithm targeting smooth optimization, the understanding of the behavior of stochastic algorithms for nonsmooth regularized optimization problems is limited, especially when the objective function is nonconvex. To fill this theoretical gap, in this paper, we propose and analyze asynchronous parallel stochastic proximal gradient (Asyn-ProxSGD) methods for nonconvex problems. We establish an ergodic convergence rate of $O(1/\sqrt{K})$ for the proposed Asyn-ProxSGD, where $K$ is the number of updates made on the model, matching the convergence rate currently known for AsynSGD (for smooth problems). To our knowledge, this is the first work that provides convergence rates of asynchronous parallel ProxSGD algorithms for nonconvex problems. Furthermore, our results are also the first to show the convergence of any stochastic proximal methods without assuming an increasing batch size or the use of additional variance reduction techniques. We implement the proposed algorithms on Parameter Server and demonstrate its convergence behavior and near-linear speedup, as the number of workers increases, on two real-world datasets.

cs.LG↗

arXiv2018

Raman Signatures of Broken Inversion Symmetry and In-plane Anisotropy in Type-II Weyl Semimetal Candidate TaIrTe4

The layered ternary compound TaIrTe4 is an important candidate to host the recently predicted type-II Weyl Fermions. However, a direct and definitive proof of the absence of inversion symmetry in this material, a prerequisite for the existence of Weyl Fermions, has so far remained evasive. Herein, an unambiguous identification of the broken inversion symmetry in TaIrTe4 is established using angle-resolved polarized Raman spectroscopy. Combining with high-resolution transmission electron microscopy, we demonstrate an efficient and nondestructive recipe to determine the exact crystallographic orientation of TaIrTe4 crystals. Such technique could be extended to the fast identification and characterization of other type-II Weyl Fermions candidates. A surprisingly strong in-plane electrical anisotropy in TaIrTe4 thin flakes is also revealed, up to 200% at 10K, which is the strongest known electrical anisotropy for materials with comparable carrier density, notably in such good metals as copper and silver.

cond-mat.mtrl-sci↗

arXiv2018

Bayesian ROC surface estimation under verification bias

The Receiver Operating Characteristic (ROC) surface is a generalization of ROC curve and is widely used for assessment of the accuracy of diagnostic tests on three categories. A complication called the verification bias, meaning that not all subjects have their true disease status verified often occur in real application of ROC analysis. This is a common problem since the gold standard test, which is used to generate true disease status, can be invasive and expensive. In this paper, we will propose a Bayesian approach for estimating the ROC surface based on continuous data under a semi-parametric trinormality assumption. Our proposed method often adopted in ROC analysis can also be extended to situation in the presence of verification bias. We compute the posterior distribution of the parameters under trinormality assumption by using a rank-based likelihood. Consistency of the posterior under mild conditions is also established. We compare our method with the existing methods for estimating ROC surface and conclude that our method performs well in terms of accuracy.

stat.AP↗

arXiv2018

A Block-wise, Asynchronous and Distributed ADMM Algorithm for General Form Consensus Optimization

Many machine learning models, including those with non-smooth regularizers, can be formulated as consensus optimization problems, which can be solved by the alternating direction method of multipliers (ADMM). Many recent efforts have been made to develop asynchronous distributed ADMM to handle large amounts of training data. However, all existing asynchronous distributed ADMM methods are based on full model updates and require locking all global model parameters to handle concurrency, which essentially serializes the updates from different workers. In this paper, we present a novel block-wise, asynchronous and distributed ADMM algorithm, which allows different blocks of model parameters to be updated in parallel. The lock-free block-wise algorithm may greatly speedup sparse optimization problems, a common scenario in reality, in which most model updates only modify a subset of all decision variables. We theoretically prove the convergence of our proposed algorithm to stationary points for non-convex general form consensus problems with possibly non-smooth regularizers. We implement the proposed ADMM algorithm on the Parameter Server framework and demonstrate its convergence and near-linear speedup performance as the number of workers increases.

cs.LG↗

arXiv2017

Learning Depth from Monocular Videos using Direct Methods

The ability to predict depth from a single image - using recent advances in CNNs - is of increasing interest to the vision community. Unsupervised strategies to learning are particularly appealing as they can utilize much larger and varied monocular video datasets during learning without the need for ground truth depth or stereo. In previous works, separate pose and depth CNN predictors had to be determined such that their joint outputs minimized the photometric error. Inspired by recent advances in direct visual odometry (DVO), we argue that the depth CNN predictor can be learned without a pose CNN predictor. Further, we demonstrate empirically that incorporation of a differentiable implementation of DVO, along with a novel depth normalization strategy - substantially improves performance over state of the art that use monocular videos for training.

cs.CV↗

arXiv2017

Semantic Photometric Bundle Adjustment on Natural Sequences

The problem of obtaining dense reconstruction of an object in a natural sequence of images has been long studied in computer vision. Classically this problem has been solved through the application of bundle adjustment (BA). More recently, excellent results have been attained through the application of photometric bundle adjustment (PBA) methods -- which directly minimize the photometric error across frames. A fundamental drawback to BA & PBA, however, is: (i) their reliance on having to view all points on the object, and (ii) for the object surface to be well textured. To circumvent these limitations we propose semantic PBA which incorporates a 3D object prior, obtained through deep learning, within the photometric bundle adjustment problem. We demonstrate state of the art performance in comparison to leading methods for object reconstruction across numerous natural sequences.

cs.CV↗