Search arXiv⌕ Search

arXiv subjects

Wei Li

Publications and source records attributed to Wei Li.

At least 1,315 records · Page 73Linked to original sources
arXiv2017

R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection

In this paper, we propose a novel method called Rotational Region CNN (R2CNN) for detecting arbitrary-oriented texts in natural scene images. The framework is based on Faster R-CNN [1] architecture. First, we use the Region Proposal Network (RPN) to generate axis-aligned bounding boxes that enclose the texts with different orientations. Second, for each axis-aligned text box proposed by RPN, we extract its pooled features with different pooled sizes and the concatenated features are used to simultaneously predict the text/non-text score, axis-aligned box and inclined minimum area box. At last, we use an inclined non-maximum suppression to get the detection results. Our approach achieves competitive results on text detection benchmarks: ICDAR 2015 and ICDAR 2013.

cs.CV↗
arXiv2017

A fast direct imaging method for the inverse obstacle scattering problem with nonlinear point scatterers

Consider the scattering of a time-harmonic plane wave by heterogeneous media consisting of linear or nonlinear point scatterers and extended obstacles. A generalized Foldy-Lax formulation is developed to take fully into account of the multiple scattering by the complex media. A new imaging function is proposed and an FFT-based direct imaging method is developed for the inverse obstacle scattering problem, which is to reconstruct the shape of the extended obstacles. The novel idea is to utilize the nonlinear point scatterers to excite high harmonic generation so that enhanced imaging resolution can be achieved. Numerical experiments are presented to demonstrate the effectiveness of the proposed method.

math.NA↗
arXiv2017

Experimental Observation of Topological Superconductivity and Majorana Zero Modes on beta-Bi2Pd Thin Films

Using a cryogenic scanning tunneling microscopy, we report the observation of topologically nontrivial superconductivity on a single material of β-Bi2Pd films grown by molecular beam epitaxy. The superconducting gap associated with spinless odd-parity pairing opens on the surface and appears much larger than the bulk one due to the Dirac-fermion enhanced parity mixing of surface pair potential. Majorana zero modes (MZMs), supported by such superconducting states, are identified at magnetic vortices. The superconductivity and MZMs exhibit resistance to nonmagnetic defects, characteristic of time-reversal-invariant topological superconductors. Our results demonstrate a simple platform to generate, manipulate and braid MZMs for quantum computation.

cond-mat.supr-con↗
arXiv2017

The q-dependent detrended cross-correlation analysis of stock market

The properties of q-dependent cross-correlation matrices of stock market have been analyzed by using the random matrix theory and complex network. The correlation structures of the fluctuations at different magnitudes have unique properties. The cross-correlations among small fluctuations are much stronger than those among large fluctuations. The large and small fluctuations are dominated by different groups of stocks. We use complex network representation to study these q-dependent matrices and discover some new identities. By utilizing those q-dependent correlation-based networks, we are able to construct some portfolio by those most independent stocks which consistently perform the best. The optimal multifractal order for portfolio optimization is approximately $q=2$. These results have deepened our understanding about the collective behaviors of the complex financial system.

q-fin.ST↗
arXiv2017

Person Re-Identification by Deep Joint Learning of Multi-Loss Classification

Existing person re-identification (re-id) methods rely mostly on either localised or global feature representation alone. This ignores their joint benefit and mutual complementary effects. In this work, we show the advantages of jointly learning local and global features in a Convolutional Neural Network (CNN) by aiming to discover correlated local and global features in different context. Specifically, we formulate a method for joint learning of local and global feature selection losses designed to optimise person re-id when using only generic matching metrics such as the L2 distance. We design a novel CNN architecture for Jointly Learning Multi-Loss (JLML) of local and global discriminative feature optimisation subject concurrently to the same re-id labelled information. Extensive comparative evaluations demonstrate the advantages of this new JLML model for person re-id over a wide range of state-of-the-art re-id methods on five benchmarks (VIPeR, GRID, CUHK01, CUHK03, Market-1501).

cs.CV↗
arXiv2017

WebVision Challenge: Visual Learning and Understanding With Web Data

We present the 2017 WebVision Challenge, a public image recognition challenge designed for deep learning based on web images without instance-level human annotation. Following the spirit of previous vision challenges, such as ILSVRC, Places2 and PASCAL VOC, which have played critical roles in the development of computer vision by contributing to the community with large scale annotated data for model designing and standardized benchmarking, we contribute with this challenge a large scale web images dataset, and a public competition with a workshop co-located with CVPR 2017. The WebVision dataset contains more than $2.4$ million web images crawled from the Internet by using queries generated from the $1,000$ semantic concepts of the benchmark ILSVRC 2012 dataset. Meta information is also included. A validation set and test set containing human annotated images are also provided to facilitate algorithmic development. The 2017 WebVision challenge consists of two tracks, the image classification task on WebVision test set, and the transfer learning task on PASCAL VOC 2012 dataset. In this paper, we describe the details of data collection and annotation, highlight the characteristics of the dataset, and introduce the evaluation metrics.

cs.CV↗
arXiv2017

Impurity- and Magnetic-field-induced Quasiparticle States in Chiral $p$-wave Superconductors

Both impurity- and magnetic-field-induced quasiparticle states in chiral $p$-wave superconductors are investigated theoretically by solving the Bogoliubov--de Gennes equations self-consistently. At the strong scattering limit, we find that a universal state bound to the impurity can be induced for both a single nonmagnetic impurity and a single magnetic impurity. Furthermore, we find that different chiral order parameters and the corresponding supercurrents have uniform distributions around linear impurities. Calculations of the local density of states in the presence of an external magnetic field show that the intensity peak of the zero-energy Majorana mode in the vortex core can be enhanced dramatically by tuning the strength of the external magnetic field or pairing interaction.

cond-mat.supr-con↗
arXiv2017

Strong anisotropy effect in iron-based superconductor CaFe$_{0.882}$Co$_{0.118}$AsF

The anisotropy of the Fe-based superconductors is much smaller than that of the cuprates and the theoretical calculations. A credible understanding for this experimental fact is still lacking up to now. Here we experimentally study the magnetic-field-angle dependence of electronic resistivity in the superconducting phase of iron-based superconductor CaFe$_{0.882}$Co$_{0.118}$AsF, and find the strongest anisotropy effect of the upper critical field among the iron-based superconductors based on the framework of Ginzburg-Landau theory. The evidences of energy band structure and charge density distribution from electronic structure calculations demonstrate that the observed strong anisotropic effect mainly comes from the strong ionic bonding in between the ions of Ca$^{2+}$ and F$^-$, which weakens the interlayer coupling between the layers of FeAs and CaF. This finding provides a significant insight into the nature of experimentally observed strong anisotropic effect of electronic resistivity, and also paves an avenue to design exotic two dimensional artificial unconventional superconductors in future.

cond-mat.supr-con↗
arXiv2017

Collective flow from AA, pA to pp collisions - Toward a unified paradigm

I give an overview of the latest development in understanding collective phenomena in high-multiplicity hadronic final state from relativistic nucleus-nucleus, proton-nucleus and proton-proton collisions. Upon reviewing the experimental data and confronting them with theoretical models, a unified paradigm in describing the observed collectivity across all hadronic collision systems is emerging. Potential future paths toward addressing key open questions, especially on collectivity in small systems (pp, pA), are discussed.

nucl-ex↗
arXiv2017

Action Unit Detection with Region Adaptation, Multi-labeling Learning and Optimal Temporal Fusing

Action Unit (AU) detection becomes essential for facial analysis. Many proposed approaches face challenging problems in dealing with the alignments of different face regions, in the effective fusion of temporal information, and in training a model for multiple AU labels. To better address these problems, we propose a deep learning framework for AU detection with region of interest (ROI) adaptation, integrated multi-label learning, and optimal LSTM-based temporal fusing. First, ROI cropping nets (ROI Nets) are designed to make sure specifically interested regions of faces are learned independently; each sub-region has a local convolutional neural network (CNN) - an ROI Net, whose convolutional filters will only be trained for the corresponding region. Second, multi-label learning is employed to integrate the outputs of those individual ROI cropping nets, which learns the inter-relationships of various AUs and acquires global features across sub-regions for AU detection. Finally, the optimal selection of multiple LSTM layers to form the best LSTM Net is carried out to best fuse temporal features, in order to make the AU prediction the most accurate. The proposed approach is evaluated on two popular AU detection datasets, BP4D and DISFA, outperforming the state of the art significantly, with an average improvement of around 13% on BP4D and 25% on DISFA, respectively.

cs.CV↗
arXiv2017

Series-Expansion Thermal Tensor Network Approach for Quantum Lattice Models

In this work we propose a series-expansion thermal tensor network (SETTN) approach for efficient simulations of quantum lattice models. This continuous-time SETTN method is based on the numerically exact Taylor series expansion of equilibrium density operator $e^{-βH}$ (with $H$ the total Hamiltonian and $β$ the imaginary time), and is thus Trotter-error free. We discover, through simulating XXZ spin chain and square-lattice quantum Ising models, that not only the Hamiltonian $H$, but also its powers $H^n$, can be efficiently expressed as matrix product operators, which enables us to calculate with high precision the equilibrium and dynamical properties of quantum lattice models at finite temperatures. Our SETTN method provides an alternative to conventional Trotter-Suzuki renormalization group (RG) approaches, and achieves an unprecedented standard of thermal RG simulations in terms of accuracy and flexibility.

cond-mat.str-el↗
arXiv2017

A CMDP-based Approach for Energy Efficient Power Allocation in Massive MIMO Systems

In this paper, energy efficient power allocation for the uplink of a multi-cell massive MIMO system is investigated. With the simplified power consumption model, the problem of power allocation is formulated as a constrained Markov decision process (CMDP) framework with infinite-horizon expected discounted total reward, which takes into account different quality of service (QoS) requirements for each user terminal (UT). We propose an offline solution containing the value iteration and Q-learning algorithms, which can obtain the global optimum power allocation policy. Simulation results show that our proposed policy performs very close to the ergodic optimal policy.

cs.IT↗
arXiv2017

Prostate Cancer Diagnosis using Deep Learning with 3D Multiparametric MRI

A novel deep learning architecture (XmasNet) based on convolutional neural networks was developed for the classification of prostate cancer lesions, using the 3D multiparametric MRI data provided by the PROSTATEx challenge. End-to-end training was performed for XmasNet, with data augmentation done through 3D rotation and slicing, in order to incorporate the 3D information of the lesion. XmasNet outperformed traditional machine learning models based on engineered features, for both train and test data. For the test data, XmasNet outperformed 69 methods from 33 participating groups and achieved the second highest AUC (0.84) in the PROSTATEx challenge. This study shows the great potential of deep learning for cancer imaging.

cs.CV↗
arXiv2017

Intra-protein binding peptide fragments have specific and intrinsic sequence patterns

The key finding in the DNA double helix model is the specific pairing or binding between nucleotides A-T and C-G, and the pairing rules are the molecule basis of genetic code. Unfortunately, no such rules have been discovered for proteins. Here we show that similar rules and intrinsic sequence patterns between intra-protein binding peptide fragments do exist, and they can be extracted using a deep learning algorithm. Multi-millions of binding and non-binding peptide fragments from currently available protein X-ray structures are classified with an accuracy of up to 93%. This discovery has the potential in helping solve protein folding and protein-protein interaction problems, two open and fundamental problems in molecular biology.

q-bio.BM↗
arXiv2017

Observation of Ten-photon Entanglement Using Thin BiB$_{3}$O$_{6}$ Crystals

We report on the experimental realization of a ten-photon Greenberger-Horne-Zeilinger state using thin BiB$_{3}$O$_{6}$ crystals. The observed fidelity is $0.606\pm0.029$, demonstrating a genuine entanglement with a standard deviation of 3.6 $σ$. This result is further verified using $p$-value calculation, obtaining an upper bound of $3.7\times10^{-3}$ under an assumed hypothesis test. Our experiment paves a new way to efficiently engineer BiB$_{3}$O$_{6}$ crystal-based multi-photon entanglement systems, which provides a promising platform for investigating advanced optical quantum information processing tasks such as boson sampling, quantum error correction and quantum-enhanced measurement.

quant-ph↗
arXiv2017

Ring-shaped potential and a class of relevant integrals involved universal associated Legendre polynomials with complicated arguments

We find that the solution of the polar angular differential equation can be written as the universal associated Legendre polynomials. Its generating function is applied to obtain an analytical result for a class of interesting integrals involving complicated argument. The present method can in principle be generalizable to the integrals involving other special functions. As an illustration we also study a typical Bessel integral with a complicated argument.

quant-ph↗
arXiv2017

Higher Spins and Yangian Symmetries

The relation between the bosonic higher spin ${\cal W}_\infty[λ]$ algebra, the affine Yangian of $\mathfrak{gl}_{1}$, and the SH$^c$ algebra is established in detail. For generic $λ$ we find explicit expressions for the low-lying ${\cal W}_\infty[λ]$ modes in terms of the affine Yangian generators, and deduce from this the precise identification between $λ$ and the parameters of the affine Yangian. Furthermore, for the free field cases corresponding to $λ=0$ and $λ=1$ we give closed-form expressions for the affine Yangian generators in terms of the free fields. Interestingly, the relation between the ${\cal W}_\infty$ modes and those of the affine Yangian is a non-local one, in general. We also establish the explicit dictionary between the affine Yangian and the SH$^c$ generators. Given that Yangian algebras are the hallmark of integrability, these identifications should pave the way towards uncovering the relation between the integrable and the higher spin symmetries.

hep-th↗