arXiv subjects

Wei Li

Publications and source records attributed to Wei Li.

At least 1,297 records · Page 72Linked to original sources

arXiv2017

Streaming Small-Footprint Keyword Spotting using Sequence-to-Sequence Models

We develop streaming keyword spotting systems using a recurrent neural network transducer (RNN-T) model: an all-neural, end-to-end trained, sequence-to-sequence model which jointly learns acoustic and language model components. Our models are trained to predict either phonemes or graphemes as subword units, thus allowing us to detect arbitrary keyword phrases, without any out-of-vocabulary words. In order to adapt the models to the requirements of keyword spotting, we propose a novel technique which biases the RNN-T system towards a specific keyword of interest. Our systems are compared against a strong sequence-trained, connectionist temporal classification (CTC) based "keyword-filler" baseline, which is augmented with a separate phoneme language model. Overall, our RNN-T system with the proposed biasing technique significantly improves performance over the baseline system.

cs.CL↗

arXiv2017

Insight-HXMT observations of the first binary neutron star merger GW170817

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.

astro-ph.HE↗

arXiv2017

Marine Wireless Big Data: Efficient Transmission, Related Applications, and Challenges

The vast volume of marine wireless sampling data and its continuously explosive growth herald the coming of the era of marine wireless big data. Two challenges imposed by these data are how to fast, reliably, and sustainably deliver them in extremely hostile marine environments and how to apply them after collection. In this article, we first propose an architecture of heterogeneous marine networks that flexibly exploits the existing underwater wireless techniques as a potential solution for fast data transmission. We then investigate the possibilities of and develop the schemes for energy-efficient and reliable undersea transmission without or slightly with data rate reduction. After discussing the data transmission, we summarize the possible applications of the collected big data and particularly focus on the problems of applying these data in sea-surface object detection and marine object recognition. Open issues and challenges that need to be further explored regarding transmission and detection/recognition are also discussed in the article.

cs.IT↗

arXiv2017

Attribute Recognition by Joint Recurrent Learning of Context and Correlation

Recognising semantic pedestrian attributes in surveillance images is a challenging task for computer vision, particularly when the imaging quality is poor with complex background clutter and uncontrolled viewing conditions, and the number of labelled training data is small. In this work, we formulate a Joint Recurrent Learning (JRL) model for exploring attribute context and correlation in order to improve attribute recognition given small sized training data with poor quality images. The JRL model learns jointly pedestrian attribute correlations in a pedestrian image and in particular their sequential ordering dependencies (latent high-order correlation) in an end-to-end encoder/decoder recurrent network. We demonstrate the performance advantage and robustness of the JRL model over a wide range of state-of-the-art deep models for pedestrian attribute recognition, multi-label image classification, and multi-person image annotation on two largest pedestrian attribute benchmarks PETA and RAP.

cs.CV↗

arXiv2017

Universal Boundary Entropies in Conformal Field Theory: A Quantum Monte Carlo Study

Recently, entropy corrections on nonorientable manifolds such as the Klein bottle are proposed as a universal characterization of critical systems with an emergent conformal field theory (CFT). We show that entropy correction on the Klein bottle can be interpreted as a boundary effect via transforming the Klein bottle into an orientable manifold with nonlocal boundary interactions. The interpretation reveals the conceptual connection of the Klein bottle entropy with the celebrated Affleck-Ludwig entropy in boundary CFT. We propose a generic scheme to extract these universal boundary entropies from quantum Monte Carlo calculation of partition function ratios in lattice models. Our numerical results on the Affleck-Ludwig entropy and Klein bottle entropy for the $q$-state quantum Potts chains with $q=2,3$ show excellent agreement with the CFT predictions. For the quantum Potts chain with $q=4$, the Klein bottle entropy slightly deviates from the CFT prediction, which is possibly due to marginally irrelevant terms in the low-energy effective theory.

cond-mat.str-el↗

arXiv2017

Hierarchical Gated Recurrent Neural Tensor Network for Answer Triggering

In this paper, we focus on the problem of answer triggering ad-dressed by Yang et al. (2015), which is a critical component for a real-world question answering system. We employ a hierarchical gated recurrent neural tensor (HGRNT) model to capture both the context information and the deep in-teractions between the candidate answers and the question. Our result on F val-ue achieves 42.6%, which surpasses the baseline by over 10 %.

cs.CL↗

arXiv2017

Dynamical transitions in a modulated Landau-Zener model with finite driving fields

We investigate a special time-dependent quantum model which assumes the Landau-Zener driving form but with an overall modulation of the intensity of the pulsing field. We demonstrate that the dynamics of the system, including the two-level case as well as its multi-level extension, is exactly solvable analytically. Differing from the original Landau-Zener model, the nonadiabatic effect of the evolution in the present driving process does not destroy the desired population transfer. As the sweep protocol employs only the finite driving fields which tend to zero asymptotically, the cutoff error due to the truncation of the driving pulse to the finite time interval turns out to be negligibly small. Furthermore, we investigate the noise effect on the driving protocol due to the dissipation of the surrounding environment. The losses of the fidelity in the protocol caused by both the phase damping process and the random spin flip noise are estimated by solving numerically the corresponding master equations within the Markovian regime.

quant-ph↗

arXiv2017

DeepUNet: A Deep Fully Convolutional Network for Pixel-level Sea-Land Segmentation

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.

cs.CV↗

arXiv2017

Conformal Thermal Tensor Network and Universal Entropy on Topological Manifolds

Partition functions of quantum critical systems, expressed as conformal thermal tensor networks, are defined on various manifolds which can give rise to universal entropy corrections. Through high-precision tensor network simulations of several quantum chains, we identify the universal entropy $S_{\mathcal{K}} = \ln{k}$ on the Klein bottle, where $k$ relates to quantum dimensions of the primary fields in conformal field theory (CFT). Different from the celebrated Affleck-Ludwig boundary entropy $\ln{g}$ ($g$ reflects non-integer groundstate degeneracy), $S_{\mathcal{K}}$ has \textit{no} boundary dependence or surface energy terms accompanied, and can be very conveniently extracted from thermal data. On the Möbius-strip manifold, we uncover an entropy $S_{\mathcal{M}} = \frac{1}{2} (\ln{g} + \ln{k})$ in CFT, where $\frac{1}{2} \ln{g}$ is associated with the only open edge of the Möbius strip, and $\frac{1}{2} \ln{k}$ with the non-orientable topology. We employ $S_{\mathcal{K}}$ to accurately pinpoint the quantum phase transitions, even for those without local order parameters.

cond-mat.str-el↗

arXiv2017

Analytic sensitivity analysis for models with correlated input variables

An analytic formula is proposed to characterize the variance propagation from correlated input variables to the model response, by using multi-variate Taylor series. With the formula, partial variance contributions to the model response are then straightforwardly evaluated in the presence of input correlations. Additionally, an arbitrary variable is represented as the sum of independent and correlated parts. Universal expressions of the coefficients that specify the correlated and independent sections of a single variable are derived by employing linear correlation model. Based on the coefficients, it is nature to quantify the independent, correlated and coupling contributions to the total variance of model response. Numerical examples suggest the effectiveness and validation of our analytic framework for general models. A practical application of the analytic framework is also proposed to the sensitivity analysis of a deterministic HIV model.

physics.soc-ph↗

arXiv2017

WebVision Database: Visual Learning and Understanding from Web Data

In this paper, we present a study on learning visual recognition models from large scale noisy web data. We build a new database called WebVision, which 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 along with those web images (e.g., title, description, tags, etc.) are also crawled. A validation set and test set containing human annotated images are also provided to facilitate algorithmic development. Based on our new database, we obtain a few interesting observations: 1) the noisy web images are sufficient for training a good deep CNN model for visual recognition; 2) the model learnt from our WebVision database exhibits comparable or even better generalization ability than the one trained from the ILSVRC 2012 dataset when being transferred to new datasets and tasks; 3) a domain adaptation issue (a.k.a., dataset bias) is observed, which means the dataset can be used as the largest benchmark dataset for visual domain adaptation. Our new WebVision database and relevant studies in this work would benefit the advance of learning state-of-the-art visual models with minimum supervision based on web data.

cs.CV↗

arXiv2017

The effects of halogen elements on the opening of an icosahedral B12 framework

The fully halogenated or hydrogenated B12X12 (X = H, F, Cl, Br and I) clusters are confirmed to be icosahedral. On the other hand, the bare B12 cluster is shown to have a planar structure. A previous study showed that a transformation from an icosahedron to a plane happens when 5 to 7 iodine atoms are substituted. Later, the transition was confirmed to be seven iodine substitutions based on an infrared spectroscopy study. In this study, we investigated the effects of different halogen atoms on the opening of the B12 icosahedral cage by means of density functional theory calculations. We found that the halogen elements do not have significant effects on the geometries of the clusters. The computed IR spectra show similar representative peaks for all halogen substituted clusters. Interestingly, we found a blue-shift in the IR spectra with the increase in the mass of the halogen atoms. Further, we compared the Gibbs free energies at different temperatures for different halogen atoms. The results show that the Gibbs free energy differences between open and close structures of B12X7 become larger when heavier halogen atoms are present. This interesting finding was subsequently investigated by energy decomposition analysis.

physics.chem-ph↗

arXiv2017

Bearing fault diagnosis under varying working condition based on domain adaptation

Traditional intelligent fault diagnosis of rolling bearings work well only under a common assumption that the labeled training data (source domain) and unlabeled testing data (target domain) are drawn from the same distribution. When the distribution changes, most fault diagnosis models need to be rebuilt from scratch using newly recollected labeled training data. However, it is expensive or impossible to annotate huge amount of training data to rebuild such new model. Meanwhile, large amounts of labeled training data have not been fully utilized yet, which is apparently a waste of resources. As one of the important research directions of transfer learning, domain adaptation (DA) typically aims at minimizing the differences between distributions of different domains in order to minimize the cross-domain prediction error by taking full advantage of information coming from both source and target domains. In this paper, we present one of the first studies on unsupervised DA in the field of fault diagnosis of rolling bearings under varying working conditions and a novel diagnosis strategy based on unsupervised DA using subspace alignment (SA) is proposed. After processed by unsupervised DA with SA, the distributions of training data and testing data become close and the classifier trained on training data can be used to classify the testing data. Experimental results on the 60 domain adaptation diagnosis problems under varying working condition in Case Western Reserve benchmark data and 12 domain adaptation diagnosis problems under varying working conditions in our new data are given to demonstrate the effectiveness of the proposed method. The proposed methods can effectively distinguish not only bearing faults categories but also fault severities.

cs.SD↗

arXiv2017

Enhanced sensitivity of low-frequency signal by using broad squeezed light and bichromatic local oscillator

We experimentally study a protocol of using the broadband high frequency squeezed vacuum to detect the low-frequency signal. In this scheme, the lower sideband field of the squeezed light carries the low frequency modulation signal and the two strong coherent light fields are applied as the bichromatic local oscillator in the homodyne detection to measure the quantum entanglement of the upper and lower sideband for the broadband squeezed light. The power of one of the local oscillators for detecting the upper sideband can be adjusted to optimize the conditional squeezing in the low frequency regime by subtracting the photocurrent of the upper sideband field of the squeezed light from that of the low sideband field. By means of the quantum correlation of the upper and lower sideband for the broadband squeezed light, the low frequency signal beyond the standard quantum limit is measured. This scheme is appropriate for enhancing sensitivity of the low frequency signal by the aid of the broad squeezed light, such as gravitational waes detection, and does not need directly produce the low frequency squeezing in the optical parametric process.

quant-ph↗

arXiv2017

SAR Target Recognition Using the Multi-aspect-aware Bidirectional LSTM Recurrent Neural Networks

The outstanding pattern recognition performance of deep learning brings new vitality to the synthetic aperture radar (SAR) automatic target recognition (ATR). However, there is a limitation in current deep learning based ATR solution that each learning process only handle one SAR image, namely learning the static scattering information, while missing the space-varying information. It is obvious that multi-aspect joint recognition introduced space-varying scattering information should improve the classification accuracy and robustness. In this paper, a novel multi-aspect-aware method is proposed to achieve this idea through the bidirectional Long Short-Term Memory (LSTM) recurrent neural networks based space-varying scattering information learning. Specifically, we first select different aspect images to generate the multi-aspect space-varying image sequences. Then, the Gabor filter and three-patch local binary pattern (TPLBP) are progressively implemented to extract a comprehensive spatial features, followed by dimensionality reduction with the Multi-layer Perceptron (MLP) network. Finally, we design a bidirectional LSTM recurrent neural network to learn the multi-aspect features with further integrating the softmax classifier to achieve target recognition. Experimental results demonstrate that the proposed method can achieve 99.9% accuracy for 10-class recognition. Besides, its anti-noise and anti-confusion performance are also better than the conventional deep learning based methods.

cs.CV↗

arXiv2017

Measurement of frequency sweep nonlinearity using atomic absorption spectroscopy

A low cost scheme to determine the frequency sweep nonlinearity using atomic saturated absorption spectroscopy is demonstrated. The frequency modulation rate is determined by directly measuring the interference fringe number and frequency gap between two atomic transition peaks of rubidium atom. Experimental results show that the frequency sweep nonlinearity is ~7.68%, with the average frequency modulation rate of ~28.95 GHz/s, which is in good agreement with theoretical expectation. With this method, the absolute optical frequency and optical path difference between two laser beams are simultaneously measured. This novel technique can be used for applications such as optical frequency sweep nonlinearity correction and real-time frequency monitor.

physics.ins-det↗

arXiv2017

BPS spectrum on AdS$_3\times $S$^3 \times $S$^3 \times $S$^1$

The BPS spectrum of string theory on AdS$_3\times {\rm S}^3 \times {\rm S}^3 \times {\rm S}^1$ is determined using a world-sheet description in terms of WZW models. It is found that the theory only has BPS states with $j^+ = j^-$ where $j^{\pm}$ refer to the spins of the $\mathfrak{su}(2)$ algebras of the large ${\cal N}=4$ superconformal algebra. We then re-examine the BPS spectrum of the corresponding supergravity and find that, in contradistinction to previous claims in the literature, also in supergravity only the states with $j^+=j^-$ are BPS. This resolves a number of long-standing puzzles regarding the BPS spectrum of string theory and supergravity in this background.

hep-th↗

arXiv2017

A holographic dual for string theory on $\mathrm{AdS}_3 \times \mathrm{S}^3 \times \mathrm{S}^3 \times \mathrm{S}^1$

The CFT dual of string theory on $\mathrm{AdS}_3 \times \mathrm{S}^3 \times \mathrm{S}^3 \times \mathrm{S}^1$ is conjectured to be the symmetric orbifold of the $\mathcal{S}_κ$ theory, provided that one of the two $Q_5^\pm$ quantum numbers is a multiple of the other. We determine the BPS spectrum of the symmetric orbifold in detail, and show that it reproduces precisely the BPS spectrum that was recently calculated in supergravity. We also determine the BPS spectrum of the world-sheet theory that is formulated in terms of WZW models, and show that, apart from some gaps (which are reminiscent of those that appear in the corresponding $\mathbb{T}^4$ calculation), it also reproduces this BPS spectrum. In fact, the matching seems to work as well as for the familiar $\mathbb{T}^4$ case, and thus our results give strong support for this proposal.

hep-th↗