arXiv subjects

Xing Liu

Publications and source records attributed to Xing Liu.

142 records · Page 8Linked to original sources

arXiv2019

Evaluating Artificial Systems for Pairwise Ranking Tasks Sensitive to Individual Differences

Owing to the advancement of deep learning, artificial systems are now rival to humans in several pattern recognition tasks, such as visual recognition of object categories. However, this is only the case with the tasks for which correct answers exist independent of human perception. There is another type of tasks for which what to predict is human perception itself, in which there are often individual differences. Then, there are no longer single "correct" answers to predict, which makes evaluation of artificial systems difficult. In this paper, focusing on pairwise ranking tasks sensitive to individual differences, we propose an evaluation method. Given a ranking result for multiple item pairs that is generated by an artificial system, our method quantifies the probability that the same ranking result will be generated by humans, and judges if it is distinguishable from human-generated results. We introduce a probabilistic model of human ranking behavior, and present an efficient computation method for the judgment. To estimate model parameters accurately from small-size samples, we present a method that uses confidence scores given by annotators for ranking each item pair. Taking as an example a task of ranking image pairs according to material attributes of objects, we demonstrate how the proposed method works.

cs.LG↗

arXiv2019

Attention-based Adaptive Selection of Operations for Image Restoration in the Presence of Unknown Combined Distortions

Many studies have been conducted so far on image restoration, the problem of restoring a clean image from its distorted version. There are many different types of distortion which affect image quality. Previous studies have focused on single types of distortion, proposing methods for removing them. However, image quality degrades due to multiple factors in the real world. Thus, depending on applications, e.g., vision for autonomous cars or surveillance cameras, we need to be able to deal with multiple combined distortions with unknown mixture ratios. For this purpose, we propose a simple yet effective layer architecture of neural networks. It performs multiple operations in parallel, which are weighted by an attention mechanism to enable selection of proper operations depending on the input. The layer can be stacked to form a deep network, which is differentiable and thus can be trained in an end-to-end fashion by gradient descent. The experimental results show that the proposed method works better than previous methods by a good margin on tasks of restoring images with multiple combined distortions.

cs.CV↗

arXiv2019

Dual Residual Networks Leveraging the Potential of Paired Operations for Image Restoration

In this paper, we study design of deep neural networks for tasks of image restoration. We propose a novel style of residual connections dubbed "dual residual connection", which exploits the potential of paired operations, e.g., up- and down-sampling or convolution with large- and small-size kernels. We design a modular block implementing this connection style; it is equipped with two containers to which arbitrary paired operations are inserted. Adopting the "unraveled" view of the residual networks proposed by Veit et al., we point out that a stack of the proposed modular blocks allows the first operation in a block interact with the second operation in any subsequent blocks. Specifying the two operations in each of the stacked blocks, we build a complete network for each individual task of image restoration. We experimentally evaluate the proposed approach on five image restoration tasks using nine datasets. The results show that the proposed networks with properly chosen paired operations outperform previous methods on almost all of the tasks and datasets.

cs.CV↗

arXiv2019

First exit and Dirichlet problem for the nonisotropic tempered $α$-stable processes

This paper discusses the first exit and Dirichlet problems of the nonisotropic tempered $α$-stable process $X_t$. The upper bounds of all moments of the first exit position $\left|X_{τ_D}\right|$ and the first exit time $τ_D$ are firstly obtained. It is found that the probability density function of $\left|X_{τ_D}\right|$ or $τ_D$ exponentially decays with the increase of $\left|X_{τ_D}\right|$ or $τ_D$, and $\mathrm{E}\left[τ_D\right]\sim \left|\mathrm{E}\left[X_{τ_D}\right]\right|$,\ $\mathrm{E}\left[τ_D\right]\sim\mathrm{E}\left[\left|X_{τ_D}-\mathrm{E}\left[X_{τ_D}\right]\right|^2\right] $. Since $\mathrmΔ^{α/2,λ}_m$ is the infinitesimal generator of the anisotropic tempered stable process, we obtain the Feynman-Kac representation of the Dirichlet problem with the operator $\mathrmΔ^{α/2,λ}_m$. Therefore, averaging the generated trajectories of the stochastic process leads to the solution of the Dirichlet problem, which is also verified by numerical experiments.

math.PR↗

arXiv2018

Data Augmentation for Robust Keyword Spotting under Playback Interference

Accurate on-device keyword spotting (KWS) with low false accept and false reject rate is crucial to customer experience for far-field voice control of conversational agents. It is particularly challenging to maintain low false reject rate in real world conditions where there is (a) ambient noise from external sources such as TV, household appliances, or other speech that is not directed at the device (b) imperfect cancellation of the audio playback from the device, resulting in residual echo, after being processed by the Acoustic Echo Cancellation (AEC) system. In this paper, we propose a data augmentation strategy to improve keyword spotting performance under these challenging conditions. The training set audio is artificially corrupted by mixing in music and TV/movie audio, at different signal to interference ratios. Our results show that we get around 30-45% relative reduction in false reject rates, at a range of false alarm rates, under audio playback from such devices.

cs.CL↗

arXiv2017

On Optimizing Distributed Tucker Decomposition for Dense Tensors

The Tucker decomposition expresses a given tensor as the product of a small core tensor and a set of factor matrices. Apart from providing data compression, the construction is useful in performing analysis such as principal component analysis (PCA)and finds applications in diverse domains such as signal processing, computer vision and text analytics. Our objective is to develop an efficient distributed implementation for the case of dense tensors. The implementation is based on the HOOI (Higher Order Orthogonal Iterator) procedure, wherein the tensor-times-matrix product forms the core routine. Prior work have proposed heuristics for reducing the computational load and communication volume incurred by the routine. We study the two metrics in a formal and systematic manner, and design strategies that are optimal under the two fundamental metrics. Our experimental evaluation on a large benchmark of tensors shows that the optimal strategies provide significant reduction in load and volume compared to prior heuristics, and provide up to 7x speed-up in the overall running time.

cs.DC↗

arXiv2017

The classical and quantum synchronization between two scattering modes in Bose-Einstein condensates generated by the standing-wave laser beams

Both classical and quantum dynamics of the synchronization between two nonlinear mechanical modes scattered from Bose-Einstein condensates (BECs) by the standing-wave laser beam are comparatively investigated. As the ultra-low dissipations of the momentum modes in the atomic BECs, the synchronized dynamics are studied in a framework of closed-system theory in order to track down both the classical and the quantum synchronizations from an angle of quantum control. The classical synchronization and the relevant dynamics of measure synchronization, the quantum synchronization and two different types of measures proposed by Mari and estimated by mutual information based on $Q$-function are studied respectively in order to reveal both the macroscopic and the microscopic signatures of synchronized behaviors in a closed quantum system. The results demonstrate that the "revival and collapse" of the quantum fluctuations beyond the classical mean-value dynamics due to long-lasting mode coherence discriminates the quantum synchronization from the classical one, which not only excludes the possibilities of an exact synchronization and a perfect density overlap in phase space, but also leads to upper limitations to Mari measure and large unceasing fluctuations to mutual information between two scattering modes. We reveal a close dynamic connection between Mari measure and the mutual information of two nonlinear momentum modes in closed BEC systems by demonstrating an opposite mean-value behavior but a similar fluctuation variation with respect to their respective evolutionary scales.

quant-ph↗

arXiv2016

Observation of electromagnetically induced Talbot effect in an atomic system

We experimentally demonstrate the Talbot effect resulting from the repeated self-reconstruction of a spatially intensity-modulated probe field under the Fresnel near-field regime. By launching the probe beam into an optically induced atomic lattice (established by interfering two coupling fields) inside a rubidium vapor cell, we can obtain an diffracted probe beam pattern from an formed electromagnetically induced grating (EIG) in a three-level $Λ$-type Doppler-free atomic configuration with the assistance of electromagnetically induced transparency (EIT). The EIG-based diffraction pattern repeats itself at the planes of integer multiple Talbot lengths, which agrees well with the theoretical prediction [Appl. Phys. Lett., 98, 081108 (2011)]. In addition, fractional EIT-induced Talbot effect was also investigated. Such experimentally demonstrated EIT Talbot effect in a coherently-prepared atomic system may pave a way for lensless and nondestructive imaging of ultracold atoms and molecules, and further demonstrating nonlinear/quantum beam dynamical features predicted for established periodic optical systems.

physics.optics↗

arXiv2016

Cosmological implications of modified gravity induced by quantum metric fluctuations

We investigate the cosmological implications of modified gravities induced by the quantum fluctuations of the gravitational metric. If the metric can be decomposed as the sum of the classical and of a fluctuating part, of quantum origin, then the corresponding Einstein quantum gravity generates at the classical level modified gravity models with a nonminimal coupling between geometry and matter. As a first step in our study, after assuming that the expectation value of the quantum correction can be generally expressed in terms of an arbitrary second order tensor constructed from the metric and from the thermodynamic quantities characterizing the matter content of the Universe, we derive the (classical) gravitational field equations in their general form. We analyze in detail the cosmological models obtained by assuming that the quantum correction tensor is given by the coupling of a scalar field and of a scalar function to the metric tensor, and by a term proportional to the matter energy-momentum tensor. For each considered model we obtain the gravitational field equations, and the generalized Friedmann equations for the case of a flat homogeneous and isotropic geometry. In some of these models the divergence of the matter energy-momentum tensor is non-zero, indicating a process of matter creation, which corresponds to an irreversible energy flow from the gravitational field to the matter fluid, and which is direct consequence of the nonminimal curvature-matter coupling. The cosmological evolution equations of these modified gravity models induced by the quantum fluctuations of the metric are investigated in detail by using both analytical and numerical methods, and it is shown that a large variety of cosmological models can be constructed, which, depending on the numerical values of the model parameters, can exhibit both accelerating and decelerating behaviors.

gr-qc↗

arXiv2016

Octave-spanning supercontinuum generation in a silicon-rich nitride waveguide

We experimentally show octave-spanning supercontinuum generation in a non-stoichiometric silicon-rich nitride waveguide when pumped by femtosecond pulses from an erbium fiber laser. The pulse energy and bandwidth are comparable to results achieved in stoichiometric silicon nitride waveguides, but our material platform is simpler to manufacture. We also observe wave-breaking supercontinuum generation by using orthogonal pumping in the same waveguide. Additional analysis reveals that the waveguide height is a powerful tuning parameter for generating mid-infrared dispersive waves while keeping the pump in the telecom band.

physics.optics↗

arXiv2016

Integrating Deep Features for Material Recognition

We propose a method for integration of features extracted using deep representations of Convolutional Neural Networks (CNNs) each of which is learned using a different image dataset of objects and materials for material recognition. Given a set of representations of multiple pre-trained CNNs, we first compute activations of features using the representations on the images to select a set of samples which are best represented by the features. Then, we measure the uncertainty of the features by computing the entropy of class distributions for each sample set. Finally, we compute the contribution of each feature to representation of classes for feature selection and integration. We examine the proposed method on three benchmark datasets for material recognition. Experimental results show that the proposed method achieves state-of-the-art performance by integrating deep features. Additionally, we introduce a new material dataset called EFMD by extending Flickr Material Database (FMD). By the employment of the EFMD with transfer learning for updating the learned CNN models, we achieve 84.0%+/-1.8% accuracy on the FMD dataset which is close to human performance that is 84.9%.

cs.CV↗

arXiv2015

Propagation dynamics of a light beam in fractional Schr\"odinger equation

Dynamics of wavepackets in fractional Schrodinger equation is still an open problem. The difficulty stems from the fact that the fractional Laplacian derivative is essentially a nonlocal operator. We investigate analytically and numerically the propagation of optical beams in fractional Schr\"odinger equation with a harmonic potential. We find that the propagation of one- and two-dimensional (1D, 2D) input chirped Gaussian beams is not harmonic. In 1D, the beam propagates along a zigzag trajectory in the real space, which corresponds to a modulated anharmonic oscillation in the momentum space. In 2D, the input Gaussian beam evolves into a breathing ring structure in both real and momentum spaces, which forms a filamented funnel-like aperiodic structure. The beams remain localized in propagation, but with increasing distance display increasingly irregular behavior, unless both the linear chirp and the transverse displacement of the incident beam are zero.

physics.optics↗

arXiv2015

Modulation of the photonic band structure topology of a honeycomb lattice in an atomic vapor

In an atomic vapor, a honeycomb lattice can be constructed by utilizing the three-beam interference method. In the method, the interference of the three beams splits the dressed energy level periodically, forming a periodic refractive index modulation with the honeycomb profile. The energy band topology of the honeycomb lattice can be modulated by frequency detunings, thereby affecting the appearance (and disappearance) of Dirac points and cones in the momentum space. This effect can be usefully exploited for the generation and manipulation of topological insulators.

physics.optics↗

arXiv2015

Self-Fourier transform and self-Fourier beams due to parabolic potential

We investigate the propagation of light beams including Hermite-Gauss, Bessel-Gauss and finite energy Airy beams in a linear medium with parabolic potential. Expectedly, the beams undergo oscillation during propagation, but quite unexpectedly they also perform automatic Fourier transform, that is, periodic change from the beam to its Fourier transform and back. The oscillating period of parity-asymmetric beams is twice that of the parity-symmetric beams. In addition to oscillation, the finite-energy Airy beams exhibit periodic inversion during propagation. Based on the propagation in parabolic potential, we introduce a class of optically-interesting beams that are self-Fourier beams -- that is, the beams whose Fourier transforms are the beams themselves.

physics.optics↗

arXiv2015

Dual accelerating Airy-Talbot recurrence effect

We demonstrate the dual accelerating Airy-Talbot recurrence effect, i.e., the self-imaging of accelerating optical beams, by propagating a superposition of Airy beams with successively changing transverse displacements. The dual Airy-Talbot effect is a spontaneous recurring imaging of the input and of the input with alternating component signs. It results from the constructive interference of Airy wave functions, which is also responsible for other kinds of Airy beams, for example, Airy breathers. An input composed of finite-energy Airy beams also displays the dual Airy-Talbot effect, but it demands a large transverse displacement and diminishes fast along the propagation direction.

physics.optics↗

arXiv2015

Mid-IR femtosecond frequency conversion by soliton-probe collision in phase-mismatched quadratic nonlinear crystals

We show numerically that ultrashort self-defocusing temporal solitons colliding with a weak pulsed probe in the near-IR can convert the probe to the mid-IR. A near-perfect conversion efficiency is possible for a high effective soliton order. The near-IR self-defocusing soliton can form in a quadratic nonlinear crystal (beta-barium borate) in the normal dispersion regime due to cascaded (phase-mismatched) second-harmonic generation, and the mid-IR converted wave is formed in the anomalous dispersion regime between $λ=2.2-2.4~μ\rm m$ as a resonant dispersive wave. This process relies on non-degenerate four-wave mixing mediated by an effective negative cross-phase modulation term caused by cascaded soliton-probe sum-frequency generation.

physics.optics↗