arXiv subjects

Yi Li

Publications and source records attributed to Yi Li.

At least 523 records · Page 29Linked to original sources

arXiv2019

DeepIM: Deep Iterative Matching for 6D Pose Estimation

Estimating the 6D pose of objects from images is an important problem in various applications such as robot manipulation and virtual reality. While direct regression of images to object poses has limited accuracy, matching rendered images of an object against the observed image can produce accurate results. In this work, we propose a novel deep neural network for 6D pose matching named DeepIM. Given an initial pose estimation, our network is able to iteratively refine the pose by matching the rendered image against the observed image. The network is trained to predict a relative pose transformation using an untangled representation of 3D location and 3D orientation and an iterative training process. Experiments on two commonly used benchmarks for 6D pose estimation demonstrate that DeepIM achieves large improvements over state-of-the-art methods. We furthermore show that DeepIM is able to match previously unseen objects.

cs.CV↗

arXiv2019

Monopole Charge Density Wave States in Weyl Semimetals

We study a new class of topological charge density wave states exhibiting monopole harmonic symmetries. The density-wave ordering is equivalent to pairing in the particle-hole channel due to Fermi surface nesting under interactions. When electron and hole Fermi surfaces carry different Chern numbers, the particle-hole pairing exhibits a non-trivial Berry phase inherited from band structure topology independent of concrete density-wave ordering mechanism. The associated density-wave gap functions become nodal, and the net nodal vorticity is determined by the monopole charge of the pairing Berry phase. The gap function nodes become zero-energy Weyl nodes of the bulk spectra of quasi-particle excitations. These states can occur in doped Weyl semimetals with nested electron and hole Fermi surfaces enclosing Weyl nodes of the same chirality in the weak coupling regime. Topologically non-trivial low-energy Fermi arc surface states appear in the density-wave ordering state as a consequence of the emergent zero-energy Weyl nodes.

cond-mat.str-el↗

arXiv2019

Oracle-Supported Dynamic Exploit Generation for Smart Contracts

Despite the high stakes involved in smart contracts, they are often developed in an undisciplined manner, leaving the security and reliability of blockchain transactions at risk. In this paper, we introduce ContraMaster: an oracle-supported dynamic exploit generation framework for smart contracts. Existing approaches mutate only single transactions; ContraMaster exceeds these by mutating the transaction sequences. ContraMaster uses data-flow, control-flow, and the dynamic contract state to guide its mutations. It then monitors the executions of target contract programs, and validates the results against a general-purpose semantic test oracle to discover vulnerabilities. Being a dynamic technique, it guarantees that each discovered vulnerability is a violation of the test oracle and is able to generate the attack script to exploit this vulnerability. In contrast to rule-based approaches, ContraMaster has not shown any false positives, and it easily generalizes to unknown types of vulnerabilities (e.g., logic errors). We evaluate ContraMaster on 218 vulnerable smart contracts. The experimental results confirm its practical applicability and advantages over the state-of-the-art techniques, and also reveal three new types of attacks.

cs.CR↗

arXiv2019

An Empirical Study on the Characteristics of Question-Answering Process on Developer Forums

Developer forums are one of the most popular and useful Q&A websites on API usages. The analysis of API forums can be a critical resource for the automated question and answer approaches. In this paper, we empirically study three API forums including Twitter, eBay, and AdWords, to investigate the characteristics of question-answering process. We observe that +60% of the posts on all three forums were answered by providing API method names or documentation. +85% of the questions were answered by API development teams and the answers from API development teams drew fewer follow-up questions. Our results provide empirical evidences for us in a future work to build automated solutions to answer developer questions on API forums.

cs.SE↗

arXiv2019

Controlled interconversion of quantized spin wave modes via local magnetic fields

In the emerging field of magnonics, spin waves are considered for information processing and transmission at high frequencies. Towards this end, the manipulation of propagating spin waves in nanostructured waveguides for novel functionality has recently been attracting increasing focus of research. Excitations with uniform magnetic fields in such waveguides favors symmetric spin wave modes with odd quantization numbers. Interference between multiple odd spin wave modes leads to a periodic self-focusing effect of the propagating spin waves. Here we demonstrate, how antisymmetric spin wave modes with even quantization numbers can be induced by local magnetic fields in a well-controlled fashion. The resulting interference patterns are discussed within an analytical model and experimentally demonstrated using microfocused Brillouin light scattering (μ-BLS).

cond-mat.mes-hall↗

arXiv2019

Data-Driven Neuron Allocation for Scale Aggregation Networks

Successful visual recognition networks benefit from aggregating information spanning from a wide range of scales. Previous research has investigated information fusion of connected layers or multiple branches in a block, seeking to strengthen the power of multi-scale representations. Despite their great successes, existing practices often allocate the neurons for each scale manually, and keep the same ratio in all aggregation blocks of an entire network, rendering suboptimal performance. In this paper, we propose to learn the neuron allocation for aggregating multi-scale information in different building blocks of a deep network. The most informative output neurons in each block are preserved while others are discarded, and thus neurons for multiple scales are competitively and adaptively allocated. Our scale aggregation network (ScaleNet) is constructed by repeating a scale aggregation (SA) block that concatenates feature maps at a wide range of scales. Feature maps for each scale are generated by a stack of downsampling, convolution and upsampling operations. The data-driven neuron allocation and SA block achieve strong representational power at the cost of considerably low computational complexity. The proposed ScaleNet, by replacing all 3x3 convolutions in ResNet with our SA blocks, achieves better performance than ResNet and its outstanding variants like ResNeXt and SE-ResNet, in the same computational complexity. On ImageNet classification, ScaleNets absolutely reduce the top-1 error rate of ResNets by 1.12 (101 layers) and 1.82 (50 layers). On COCO object detection, ScaleNets absolutely improve the mmAP with backbone of ResNets by 3.6 (101 layers) and 4.6 (50 layers) on Faster RCNN, respectively. Code and models are released at https://github.com/Eli-YiLi/ScaleNet.

cs.CV↗

arXiv2019

REPAIR: Removing Representation Bias by Dataset Resampling

Modern machine learning datasets can have biases for certain representations that are leveraged by algorithms to achieve high performance without learning to solve the underlying task. This problem is referred to as "representation bias". The question of how to reduce the representation biases of a dataset is investigated and a new dataset REPresentAtion bIas Removal (REPAIR) procedure is proposed. This formulates bias minimization as an optimization problem, seeking a weight distribution that penalizes examples easy for a classifier built on a given feature representation. Bias reduction is then equated to maximizing the ratio between the classification loss on the reweighted dataset and the uncertainty of the ground-truth class labels. This is a minimax problem that REPAIR solves by alternatingly updating classifier parameters and dataset resampling weights, using stochastic gradient descent. An experimental set-up is also introduced to measure the bias of any dataset for a given representation, and the impact of this bias on the performance of recognition models. Experiments with synthetic and action recognition data show that dataset REPAIR can significantly reduce representation bias, and lead to improved generalization of models trained on REPAIRed datasets. The tools used for characterizing representation bias, and the proposed dataset REPAIR algorithm, are available at https://github.com/JerryYLi/Dataset-REPAIR/.

cs.CV↗

arXiv2019

Biphasic Learning of GANs for High-Resolution Image-to-Image Translation

Despite that the performance of image-to-image translation has been significantly improved by recent progress in generative models, current methods still suffer from severe degradation in training stability and sample quality when applied to the high-resolution situation. In this work, we present a novel training framework for GANs, namely biphasic learning, to achieve image-to-image translation in multiple visual domains at $1024^2$ resolution. Our core idea is to design an adjustable objective function that varies across training phases. Within the biphasic learning framework, we propose a novel inherited adversarial loss to achieve the enhancement of model capacity and stabilize the training phase transition. Furthermore, we introduce a perceptual-level consistency loss through mutual information estimation and maximization. To verify the superiority of the proposed method, we apply it to a wide range of face-related synthesis tasks and conduct experiments on multiple large-scale datasets. Through comprehensive quantitative analyses, we demonstrate that our method significantly outperforms existing methods.

cs.CV↗

arXiv2019

Nutation spectroscopy of a nanomagnet driven into deeply nonlinear ferromagnetic resonance

Strongly out-of-equilibrium regimes in magnetic nanostructures exhibit novel properties, linked to the nonlinear nature of magnetization dynamics, which are of great fundamental and practical interest. Here, we demonstrate that field-driven ferromagnetic resonance can occur with substantial spatial coherency at unprecedented large angle of magnetization precessions, which is normally prevented by the onset of spin-wave instabilities and magnetization turbulent dynamics. Our results show that this limitation can be overcome in nanomagnets, where the geometric confinement drastically reduces the density of spin-wave modes. The obtained deeply nonlinear ferromagnetic resonance regime is probed by a new spectroscopic technique based on the application of a second excitation field. This enables to resonantly drive slow coherent magnetization nutations around the large angle periodic trajectory. Our experimental findings are well accounted for by an analytical model derived for systems with uniaxial symmetry. They also provide new means for controlling highly nonlinear magnetization dynamics in nanostructures, which open interesting applicative opportunities in the context of magnetic nanotechnologies.

cond-mat.mes-hall↗

arXiv2019

Strong magnon-photon coupling in ferromagnet-superconducting resonator thin-film devices

We demonstrate strong magnon-photon coupling of a thin-film permalloy device fabricated on a coplanar superconducting resonator. A coupling strength of 0.152 GHz and a cooperativity of 68 are found for a 30-nm-thick permalloy stripe. The coupling strength is tunable by rotating the biasing magnetic field or changing the volume of permalloy. We also observe an enhancement of magnon-photon coupling in the nonlinear regime of the superconducting resonator, which is mediated by the nucleation of dynamic flux vortices. Our results demonstrate a critical step towards future integrated hybrid systems for quantum magnonics and on-chip coherent information transfer.

cond-mat.mes-hall↗

arXiv2019

Detecting Lesion Bounding Ellipses With Gaussian Proposal Networks

Lesions characterized by computed tomography (CT) scans, are arguably often elliptical objects. However, current lesion detection systems are predominantly adopted from the popular Region Proposal Networks (RPNs) that only propose bounding boxes without fully leveraging the elliptical geometry of lesions. In this paper, we present Gaussian Proposal Networks (GPNs), a novel extension to RPNs, to detect lesion bounding ellipses. Instead of directly regressing the rotation angle of the ellipse as the common practice, GPN represents bounding ellipses as 2D Gaussian distributions on the image plain and minimizes the Kullback-Leibler (KL) divergence between the proposed Gaussian and the ground truth Gaussian for object localization. We show the KL divergence loss approximately incarnates the regression loss in the RPN framework when the rotation angle is 0. Experiments on the DeepLesion dataset show that GPN significantly outperforms RPN for lesion bounding ellipse detection thanks to lower localization error. GPN is open sourced at https://github.com/baidu-research/GPN

cs.CV↗

arXiv2019

Direct observation of vacuum arc evolution with nanosecond resolution

Sufficiently high voltage applied between two metal electrodes, even in ultra high vacuum conditions, results in an inevitable discharge that lights up the entire gap, opening a conductive channel through the vacuum and parasitically consuming large amounts of energy. Despite many efforts to understand the processes that lead to this phenomenon, known as vacuum arc, there is still no consensus regarding the role of each electrode in the evolution of such a momentous process as lightning. Employing a high-speed camera, we capture the entire lightning process step-by-step with a nanosecond resolution and find which of the two electrodes holds the main responsibility for igniting the arc. The light that gradually expands from the positively charged electrode (anode), often is assumed to play the main role in the formation of a vacuum arc. However, both the nanosecond-resolution images of vacuum arc evolution and the corresponding theoretical calculations agree that the conductive channel between the electrodes is built in the form of cathodic plasma long before any significant activity develops in the anode. We show evidently that the anode illumination is weaker and plays a minor role in igniting and maintaining the conductive channel.

physics.app-ph↗

arXiv2019

Simultaneous Optical and Electrical Spin-Torque Magnetometry with Stroboscopic Detection of Spin-Precession Phase

Spin-based coherent information processing and encoding utilize the precession phase of spins in magnetic materials. However, the detection and manipulation of spin precession phases remain a major challenge for advanced spintronic functionalities. By using simultaneous electrical and optical detection, we demonstrate the direct measurement of the precession phase of Permalloy ferromagnetic resonance driven by the spin-orbit torques from adjacent heavy metals. The spin Hall angle of the heavy metals can be independently determined from concurrent electrical and optical signals. The stroboscopic optical detection also allows spatially measuring local spin-torque parameters and the induced ferromagnetic resonance with comprehensive amplitude and phase information. Our study offers a route towards future advanced characterizations of spin-torque oscillators, magnonic circuits, and tunnelling junctions, where measuring the current-induced spin dynamics of individual nanomagnets are required.

cond-mat.mes-hall↗

arXiv2019

Giant anisotropy of Gilbert damping in epitaxial CoFe films

Tailoring Gilbert damping of metallic ferromagnetic thin films is one of the central interests in spintronics applications. Here we report a giant Gilbert damping anisotropy in epitaxial Co$_{50}$Fe$_{50}$ thin film with a maximum-minimum damping ratio of 400 \%, determined by broadband spin-torque as well as inductive ferromagnetic resonance. We conclude that the origin of this damping anisotropy is the variation of the spin orbit coupling for different magnetization orientations in the cubic lattice, which is further corroborate from the magnitude of the anisotropic magnetoresistance in Co$_{50}$Fe$_{50}$.

cond-mat.mtrl-sci↗

arXiv2018

DeepCruiser: Automated Guided Testing for Stateful Deep Learning Systems

Deep learning (DL) defines a data-driven programming paradigm that automatically composes the system decision logic from the training data. In company with the data explosion and hardware acceleration during the past decade, DL achieves tremendous success in many cutting-edge applications. However, even the state-of-the-art DL systems still suffer from quality and reliability issues. It was only until recently that some preliminary progress was made in testing feed-forward DL systems. In contrast to feed-forward DL systems, recurrent neural networks (RNN) follow a very different architectural design, implementing temporal behaviors and memory with loops and internal states. Such stateful nature of RNN contributes to its success in handling sequential inputs such as audio, natural languages and video processing, but also poses new challenges for quality assurance. In this paper, we initiate the very first step towards testing RNN-based stateful DL systems. We model RNN as an abstract state transition system, based on which we define a set of test coverage criteria specialized for stateful DL systems. Moreover, we propose an automated testing framework, DeepCruiser, which systematically generates tests in large scale to uncover defects of stateful DL systems with coverage guidance. Our in-depth evaluation on a state-of-the-art speech-to-text DL system demonstrates the effectiveness of our technique in improving quality and reliability of stateful DL systems.

cs.SE↗

arXiv2018

Equivalence of Coefficients Extraction of One-loop Master Integrals

Now there are many different methods to do the PV-reduction for the one loop amplitudes. Two of them are unitarity cut method and generalized unitarity cut method. In this short paper, we present an explicit connection of these two methods, especially how the extractions of triangle and bubble coefficients are equivalent to each other.

hep-th↗

arXiv2018

Forecasting Transportation Network Speed Using Deep Capsule Networks with Nested LSTM Models

Accurate and reliable traffic forecasting for complicated transportation networks is of vital importance to modern transportation management. The complicated spatial dependencies of roadway links and the dynamic temporal patterns of traffic states make it particularly challenging. To address these challenges, we propose a new capsule network (CapsNet) to extract the spatial features of traffic networks and utilize a nested LSTM (NLSTM) structure to capture the hierarchical temporal dependencies in traffic sequence data. A framework for network-level traffic forecasting is also proposed by sequentially connecting CapsNet and NLSTM. On the basis of literature review, our study is the first to adopt CapsNet and NLSTM in the field of traffic forecasting. An experiment on a Beijing transportation network with 278 links shows that the proposed framework with the capability of capturing complicated spatiotemporal traffic patterns outperforms multiple state-of-the-art traffic forecasting baseline models. The superiority and feasibility of CapsNet and NLSTM are also demonstrated, respectively, by visualizing and quantitatively evaluating the experimental results.

cs.LG↗

arXiv2018

Image Super-Resolution Using TV Priori Guided Convolutional Network

We proposed a TV priori information guided deep learning method for single image super-resolution(SR). The new alogorithm up-sample method based on TV priori, new learning method and neural networks architecture are embraced in our TV guided priori Convolutional Neural Network which diretcly learns an end to end mapping between the low level to high level images.

cs.MM↗