arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 1,423 records · Page 79Linked to original sources

arXiv2020

Outer scale of the wide-range Prandtl/Schmidt number spectrum on beam wander for oceanic optical turbulence

Light propagation in ocean is influenced by the refractive-index which is related to temperature, salinity, outer-scale, etc. Based on Hill's model 1 (H1), two kinds of oceanic refractive-index spectrum (ORIS) have been proposed to describe the second order characteristic of refractive-index. Most recently, several ORIS models were proposed based on Hill's model 4 (H4), which gave a better precision in high wave-numbers (viscous-diffusive range). However, the outer scale, as a key parameter related to practical environment, has not been introduced into any oceanic H4-based spectra. In this paper, we take the outer-scale parameter into an H4-based spectrum which is adapted to the wide-range Prandtl/Schmidt number [Opt. Express. 20, 11111(2019)]. The proposed outer-scaled spectrum could be used in analyzing wave propagation in limited outer-scaled environment with different values of average temperature and salinity. We further derived the beam wander formula of collimated laser beam. Numerical calculations show that the beam wander influenced by outer-scale length $L_{0}$ is more obvious than that influenced by average temperature $\langle T\rangle$, when $L_{0}$ varies from $10 \rm m$ to $100 \rm m$, and $\langle T\rangle$ ranges from $0^{\circ} \rm C$ to $30^{\circ} \rm C$. When salinity fluctuations prevails ($ω\rightarrow 0$), the influence of outer scale becomes weaker. In contribution proportion of beam wander, the temperature-salt coupling term is the much larger than that of temperature or salinity term.

physics.ao-ph↗

arXiv2019

Experimental evidence for orbital magnetic moments generated by moiré-scale current loops in twisted bilayer graphene

A remarkable property of twisted bilayer graphene (TBG) with small twist angle is the presence of a well-defined and conserved low-energy valley degrees of freedom1, which can potentially bring about new types of valley-associated spontaneous-symmetry breaking phases. Electron-electron (e-e) interactions in the TBG near the magic angle 1.1 degree can lift the valley degeneracy, allowing for the realization of orbital magnetism and topological phases2-11. However, direct measurement of the orbital-based magnetism in the TBG is still lacking up to now. Here we report evidence for orbital magnetic moment generated by the moire-scale current loops in a TBG with a twist angle θ ~ 1.68 degree. The valley degeneracy of the 1.68 degree TBG is removed by e-e interactions when its low-energy van Hove singularity (VHS) is nearly half filled. A large and linear response of the valley splitting to magnetic fields is observed, attributing to coupling to the large orbital magnetic moment induced by chiral current loops circulating in the moire pattern. According to our experiment, the orbital magnetic moment is about 10.7 uB per moire supercell. Our result paves the way to explore magnetism that is purely orbital in slightly twisted graphene system.

cond-mat.mtrl-sci↗

arXiv2019

A Nonlinear Regression Method for Composite Protection Modeling of Induction Motor Loads

Protection equipment is used to prevent damage to induction motor loads by isolating them from power systems in the event of severe faults. Modeling the response of induction motor loads and their protection is vital for power system planning and operation, especially in understanding system's dynamic performance and stability after a fault occurs. Induction motors are usually equipped with several types of protection with different operation mechanisms, making it challenging to develop adequate yet not overly complex protection models and determine their parameters for aggregate induction motor models. This paper proposes an optimization-based nonlinear regression framework to determine protection model parameters for aggregate induction motor loads in commercial buildings. Using a mathematical abstraction, the task of determining a suitable set of parameters for the protection model in composite load models is formulated as a nonlinear regression problem. Numerical examples are provided to illustrate the application of the framework. Sensitivity studies are presented to demonstrate the impact of lack of available motor load information on the accuracy of the protection models.

eess.SY↗

arXiv2019

Diagnosing the Stellar Population and Tidal Structure of the Blanco1 Star Cluster

We present the stellar population, using {\it Gaia}\,DR2 parallax, kinematics, and photometry, of the young ($\sim 100$~Myr), nearby ($\sim 230$~pc) open cluster, Blanco1. A total of 644 member candidates are identified via the unsupervised machine learning method \textsc{StarGO} to find the clustering in the 5-dimensional position and proper motion parameter ($X$, $Y$, $Z$, $μ_α\cosδ$, $μ_δ$) space. Within the tidal radius of $10.0 \pm 0.3$~pc, there are 488 member candidates, 3 times more than those outside. A leading tail and a trailing tail, each of 50--60~pc in the Galactic plane, are found for the first time for this cluster, with stars further from the cluster center streaming away faster, manifest stellar stripping. Blanco1 has a total detected mass of $285\pm32$~M$_\odot $ with a mass function consistent with a slope of $α=1.35\pm0.2$ in the sense of $dN/dm \propto m^{-α}$, in the mass range of 0.25--2.51~M$_\odot $, where $N$ is the number of members and $m$ is stellar mass. A Minimum Spanning Tree ($Λ_{\rm MSR}$) analysis shows the cluster to be moderately mass segregated among the most massive members ($\gtrsim 1.4$~M$_\odot$), suggesting an early stage of dynamical disintegration.

astro-ph.GA↗

arXiv2019

SpecAugment on Large Scale Datasets

Recently, SpecAugment, an augmentation scheme for automatic speech recognition that acts directly on the spectrogram of input utterances, has shown to be highly effective in enhancing the performance of end-to-end networks on public datasets. In this paper, we demonstrate its effectiveness on tasks with large scale datasets by investigating its application to the Google Multidomain Dataset (Narayanan et al., 2018). We achieve improvement across all test domains by mixing raw training data augmented with SpecAugment and noise-perturbed training data when training the acoustic model. We also introduce a modification of SpecAugment that adapts the time mask size and/or multiplicity depending on the length of the utterance, which can potentially benefit large scale tasks. By using adaptive masking, we are able to further improve the performance of the Listen, Attend and Spell model on LibriSpeech to 2.2% WER on test-clean and 5.2% WER on test-other.

eess.AS↗

arXiv2019

SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition

We present SpecAugment, a simple data augmentation method for speech recognition. SpecAugment is applied directly to the feature inputs of a neural network (i.e., filter bank coefficients). The augmentation policy consists of warping the features, masking blocks of frequency channels, and masking blocks of time steps. We apply SpecAugment on Listen, Attend and Spell networks for end-to-end speech recognition tasks. We achieve state-of-the-art performance on the LibriSpeech 960h and Swichboard 300h tasks, outperforming all prior work. On LibriSpeech, we achieve 6.8% WER on test-other without the use of a language model, and 5.8% WER with shallow fusion with a language model. This compares to the previous state-of-the-art hybrid system of 7.5% WER. For Switchboard, we achieve 7.2%/14.6% on the Switchboard/CallHome portion of the Hub5'00 test set without the use of a language model, and 6.8%/14.1% with shallow fusion, which compares to the previous state-of-the-art hybrid system at 8.3%/17.3% WER.

eess.AS↗

arXiv2019

Neural Recovery Machine for Chinese Dropped Pronoun

Dropped pronouns (DPs) are ubiquitous in pro-drop languages like Chinese, Japanese etc. Previous work mainly focused on painstakingly exploring the empirical features for DPs recovery. In this paper, we propose a neural recovery machine (NRM) to model and recover DPs in Chinese, so that to avoid the non-trivial feature engineering process. The experimental results show that the proposed NRM significantly outperforms the state-of-the-art approaches on both two heterogeneous datasets. Further experiment results of Chinese zero pronoun (ZP) resolution show that the performance of ZP resolution can also be improved by recovering the ZPs to DPs.

cs.CL↗

arXiv2019

FeCaffe: FPGA-enabled Caffe with OpenCL for Deep Learning Training and Inference on Intel Stratix 10

Deep learning and Convolutional Neural Network (CNN) have becoming increasingly more popular and important in both academic and industrial areas in recent years cause they are able to provide better accuracy and result in classification, detection and recognition areas, compared to traditional approaches. Currently, there are many popular frameworks in the market for deep learning development, such as Caffe, TensorFlow, Pytorch, and most of frameworks natively support CPU and consider GPU as the mainline accelerator by default. FPGA device, viewed as a potential heterogeneous platform, still cannot provide a comprehensive support for CNN development in popular frameworks, in particular to the training phase. In this paper, we firstly propose the FeCaffe, i.e. FPGA-enabled Caffe, a hierarchical software and hardware design methodology based on the Caffe to enable FPGA to support mainline deep learning development features, e.g. training and inference with Caffe. Furthermore, we provide some benchmarks with FeCaffe by taking some classical CNN networks as examples, and further analysis of kernel execution time in details accordingly. Finally, some optimization directions including FPGA kernel design, system pipeline, network architecture, user case application and heterogeneous platform levels, have been proposed gradually to improve FeCaffe performance and efficiency. The result demonstrates the proposed FeCaffe is capable of supporting almost full features during CNN network training and inference respectively with high degree of design flexibility, expansibility and reusability for deep learning development. Compared to prior studies, our architecture can support more network and training settings, and current configuration can achieve 6.4x and 8.4x average execution time improvement for forward and backward respectively for LeNet.

cs.DC↗

arXiv2019

Fault Location in Power Distribution Systems via Deep Graph Convolutional Networks

This paper develops a novel graph convolutional network (GCN) framework for fault location in power distribution networks. The proposed approach integrates multiple measurements at different buses while taking system topology into account. The effectiveness of the GCN model is corroborated by the IEEE 123 bus benchmark system. Simulation results show that the GCN model significantly outperforms other widely-used machine learning schemes with very high fault location accuracy. In addition, the proposed approach is robust to measurement noise and data loss errors. Data visualization results of two competing neural networks are presented to explore the mechanism of GCN's superior performance. A data augmentation procedure is proposed to increase the robustness of the model under various levels of noise and data loss errors. Further experiments show that the model can adapt to topology changes of distribution networks and perform well with a limited number of measured buses.

cs.LG↗

arXiv2019

Scale- and Context-Aware Convolutional Non-intrusive Load Monitoring

Non-intrusive load monitoring addresses the challenging task of decomposing the aggregate signal of a household's electricity consumption into appliance-level data without installing dedicated meters. By detecting load malfunction and recommending energy reduction programs, cost-effective non-intrusive load monitoring provides intelligent demand-side management for utilities and end users. In this paper, we boost the accuracy of energy disaggregation with a novel neural network structure named scale- and context-aware network, which exploits multi-scale features and contextual information. Specifically, we develop a multi-branch architecture with multiple receptive field sizes and branch-wise gates that connect the branches in the sub-networks. We build a self-attention module to facilitate the integration of global context, and we incorporate an adversarial loss and on-state augmentation to further improve the model's performance. Extensive simulation results tested on open datasets corroborate the merits of the proposed approach, which significantly outperforms state-of-the-art methods.

eess.SP↗

arXiv2019

A comparison of end-to-end models for long-form speech recognition

End-to-end automatic speech recognition (ASR) models, including both attention-based models and the recurrent neural network transducer (RNN-T), have shown superior performance compared to conventional systems. However, previous studies have focused primarily on short utterances that typically last for just a few seconds or, at most, a few tens of seconds. Whether such architectures are practical on long utterances that last from minutes to hours remains an open question. In this paper, we both investigate and improve the performance of end-to-end models on long-form transcription. We first present an empirical comparison of different end-to-end models on a real world long-form task and demonstrate that the RNN-T model is much more robust than attention-based systems in this regime. We next explore two improvements to attention-based systems that significantly improve its performance: restricting the attention to be monotonic, and applying a novel decoding algorithm that breaks long utterances into shorter overlapping segments. Combining these two improvements, we show that attention-based end-to-end models can be very competitive to RNN-T on long-form speech recognition.

eess.AS↗

arXiv2019

Transferable End-to-End Aspect-based Sentiment Analysis with Selective Adversarial Learning

Joint extraction of aspects and sentiments can be effectively formulated as a sequence labeling problem. However, such formulation hinders the effectiveness of supervised methods due to the lack of annotated sequence data in many domains. To address this issue, we firstly explore an unsupervised domain adaptation setting for this task. Prior work can only use common syntactic relations between aspect and opinion words to bridge the domain gaps, which highly relies on external linguistic resources. To resolve it, we propose a novel Selective Adversarial Learning (SAL) method to align the inferred correlation vectors that automatically capture their latent relations. The SAL method can dynamically learn an alignment weight for each word such that more important words can possess higher alignment weights to achieve fine-grained (word-level) adaptation. Empirically, extensive experiments demonstrate the effectiveness of the proposed SAL method.

cs.CL↗

arXiv2019

Solving Optimization Problems through Fully Convolutional Networks: an Application to the Travelling Salesman Problem

In the new wave of artificial intelligence, deep learning is impacting various industries. As a closely related area, optimization algorithms greatly contribute to the development of deep learning. But the reverse applications are still insufficient. Is there any efficient way to solve certain optimization problem through deep learning? The key is to convert the optimization to a representation suitable for deep learning. In this paper, a traveling salesman problem (TSP) is studied. Considering that deep learning is good at image processing, an image representation method is proposed to transfer a TSP to an image. Based on samples of a 10 city TSP, a fully convolutional network (FCN) is used to learn the mapping from a feasible region to an optimal solution. The training process is analyzed and interpreted through stages. A visualization method is presented to show how a FCN can understand the training task of a TSP. Once the training is completed, no significant effort is required to solve a new TSP and the prediction is obtained on the scale of milliseconds. The results show good performance in finding the global optimal solution. Moreover, the developed FCN model has been demonstrated on TSP's with different city numbers, proving excellent generalization performance.

cs.LG↗

arXiv2019

End-to-End Multi-View Fusion for 3D Object Detection in LiDAR Point Clouds

Recent work on 3D object detection advocates point cloud voxelization in birds-eye view, where objects preserve their physical dimensions and are naturally separable. When represented in this view, however, point clouds are sparse and have highly variable point density, which may cause detectors difficulties in detecting distant or small objects (pedestrians, traffic signs, etc.). On the other hand, perspective view provides dense observations, which could allow more favorable feature encoding for such cases. In this paper, we aim to synergize the birds-eye view and the perspective view and propose a novel end-to-end multi-view fusion (MVF) algorithm, which can effectively learn to utilize the complementary information from both. Specifically, we introduce dynamic voxelization, which has four merits compared to existing voxelization methods, i) removing the need of pre-allocating a tensor with fixed size; ii) overcoming the information loss due to stochastic point/voxel dropout; iii) yielding deterministic voxel embeddings and more stable detection outcomes; iv) establishing the bi-directional relationship between points and voxels, which potentially lays a natural foundation for cross-view feature fusion. By employing dynamic voxelization, the proposed feature fusion architecture enables each point to learn to fuse context information from different views. MVF operates on points and can be naturally extended to other approaches using LiDAR point clouds. We evaluate our MVF model extensively on the newly released Waymo Open Dataset and on the KITTI dataset and demonstrate that it significantly improves detection accuracy over the comparable single-view PointPillars baseline.

cs.CV↗

arXiv2019

One-step implementation of a multi-target-qubit controlled phase gate with cat-state qubits in circuit QED

We propose a single-step implementation of a muti-target-qubit controlled phase gate with one cat-state qubit (\textit{cqubit}) simultaneously controlling $n-1$ target \textit{cqubits}. The two logic states of a \textit{cqubit} are represented by two orthogonal cat states of a single cavity mode. In this proposal, the gate is implemented with $n$ microwave cavities coupled to a superconducting transmon qutrit. Because the qutrit remains in the ground state during the gate operation, decoherence caused due to the qutrit's energy relaxation and dephasing is greatly suppressed. The gate implementation is quite simple because only a single-step operation is needed and neither classical pulse nor measurement is required. Numerical simulations demonstrate that high-fidelity realization of a controlled phase gate with one cqubit simultaneously controlling two target cqubits is feasible with present circuit QED technology. This proposal can be extended to a wide range of physical systems to realize the proposed gate, such as multiple microwave or optical cavities coupled to a natural or artificial three-level atom.

quant-ph↗

arXiv2019

Circuit QED: single-step realization of a multiqubit controlled phase gate with one microwave photonic qubit simultaneously controlling $n-1$ microwave photonic qubits

We present a novel method to realize a multi-target-qubit controlled phase gate with one microwave photonic qubit simultaneously controlling $n-1$ target microwave photonic qubits. This gate is implemented with $n$ microwave cavities coupled to a superconducting flux qutrit. Each cavity hosts a microwave photonic qubit, whose two logic states are represented by the vacuum state and the single photon state of a single cavity mode, respectively. During the gate operation, the qutrit remains in the ground state and thus decoherence from the qutrit is greatly suppressed. This proposal requires only a single-step operation and thus the gate implementation is quite simple. The gate operation time is independent of the number of the qubits. In addition, this proposal does not need applying classical pulse or any measurement. Numerical simulations demonstrate that high-fidelity realization of a controlled phase gate with one microwave photonic qubit simultaneously controlling two target microwave photonic qubits is feasible with current circuit QED technology. The proposal is quite general and can be applied to implement the proposed gate in a wide range of physical systems, such as multiple microwave or optical cavities coupled to a natural or artificial $Λ$-type three-level atom.

quant-ph↗

arXiv2019

Universal quantum gate with hybrid qubits in circuit quantum electrodynamics

Hybrid qubits have recently drawn intensive attention in quantum computing. We here propose a method to implement a universal controlled-phase gate of two hybrid qubits via two three-dimensional (3D) microwave cavities coupled to a superconducting flux qutrit. For the gate considered here, the control qubit is a microwave photonic qubit (particle-like qubit), whose two logic states are encoded by the vacuum state and the single-photon state of a cavity, while the target qubit is a cat-state qubit (wave-like qubit), whose two logic states are encoded by the two orthogonal cat states of the other cavity. During the gate operation, the qutrit remains in the ground state; therefore decoherence from the qutrit is greatly suppressed. The gate realization is quite simple, because only a single basic operation is employed and neither classical pulse nor measurement is used. Our numerical simulations demonstrate that with current circuit QED technology, this gate can be realized with a high fidelity. The generality of this proposal allows to implement the proposed gate in a wide range of physical systems, such as two 1D or 3D microwave or optical cavities coupled to a natural or artificial three-level atom. Finally, this proposal can be applied to create a novel entangled state between a particle-like photonic qubit and a wave-like cat-state qubit.

quant-ph↗

arXiv2019

Orbital motion of test particles in regular Hayward black hole space-time

In this paper, all possible orbits of test particles are investigated by using phase plane method in regular Hayward black hole space-time. Our results show that the time-like orbits are divided into four types: unstable circular orbits, separates stable orbits, stable hyperbolic orbits and elliptical orbits in regular Hayward black hole space-time. We find that the orbital properties vary with the change of $\ell$ (a convenient encoding of the central energy density $3/8π\ell^{2}$). If $\ell =\frac{1}{3}$ and $b < 3.45321$, the test particles which moving toward the black hole will definitely be plunging into the black hole. In addition, it is obtained that the innermost stable circular orbit happens at $r_{min}$ = 5.93055 for $b$ = 3.45321.

gr-qc↗