arXiv subjects

Yang Wang

Publications and source records attributed to Yang Wang.

At least 793 records · Page 44Linked to original sources

arXiv2018

Cycle-Consistent Deep Generative Hashing for Cross-Modal Retrieval

In this paper, we propose a novel deep generative approach to cross-modal retrieval to learn hash functions in the absence of paired training samples through the cycle consistency loss. Our proposed approach employs adversarial training scheme to lean a couple of hash functions enabling translation between modalities while assuming the underlying semantic relationship. To induce the hash codes with semantics to the input-output pair, cycle consistency loss is further proposed upon the adversarial training to strengthen the correlations between inputs and corresponding outputs. Our approach is generative to learn hash functions such that the learned hash codes can maximally correlate each input-output correspondence, meanwhile can also regenerate the inputs so as to minimize the information loss. The learning to hash embedding is thus performed to jointly optimize the parameters of the hash functions across modalities as well as the associated generative models. Extensive experiments on a variety of large-scale cross-modal data sets demonstrate that our proposed method achieves better retrieval results than the state-of-the-arts.

cs.CV↗

arXiv2018

Video-based Person Re-identification Using Spatial-Temporal Attention Networks

We consider the problem of video-based person re-identification. The goal is to identify a person from videos captured under different cameras. In this paper, we propose an efficient spatial-temporal attention based model for person re-identification from videos. Our method generates an attention score for each frame based on frame-level features. The attention scores of all frames in a video are used to produce a weighted feature vector for the input video. Unlike most existing deep learning methods that use global representation, our approach focuses on attention scores. Extensive experiments on two benchmark datasets demonstrate that our method achieves the state-of-the-art performance. This is a technical report.

cs.CV↗

arXiv2018

Data-Driven Tight Frame for Cryo-EM Image Denoising and Conformational Classification

The cryo-electron microscope (cryo-EM) is increasingly popular these years. It helps to uncover the biological structures and functions of macromolecules. In this paper, we address image denoising problem in cryo-EM. Denoising the cryo-EM images can help to distinguish different molecular conformations and improve three dimensional reconstruction resolution. We introduce the use of data-driven tight frame (DDTF) algorithm for cryo-EM image denoising. The DDTF algorithm is closely related to the dictionary learning. The advantage of DDTF algorithm is that it is computationally efficient, and can well identify the texture and shape of images without using large data samples. Experimental results on cryo-EM image denoising and conformational classification demonstrate the power of DDTF algorithm for cryo-EM image denoising and classification.

stat.CO↗

arXiv2018

Perceptual Visual Interactive Learning

Supervised learning methods are widely used in machine learning. However, the lack of labels in existing data limits the application of these technologies. Visual interactive learning (VIL) compared with computers can avoid semantic gap, and solve the labeling problem of small label quantity (SLQ) samples in a groundbreaking way. In order to fully understand the importance of VIL to the interaction process, we re-summarize the interactive learning related algorithms (e.g. clustering, classification, retrieval etc.) from the perspective of VIL. Note that, perception and cognition are two main visual processes of VIL. On this basis, we propose a perceptual visual interactive learning (PVIL) framework, which adopts gestalt principle to design interaction strategy and multi-dimensionality reduction (MDR) to optimize the process of visualization. The advantage of PVIL framework is that it combines computer's sensitivity of detailed features and human's overall understanding of global tasks. Experimental results validate that the framework is superior to traditional computer labeling methods (such as label propagation) in both accuracy and efficiency, which achieves significant classification results on dense distribution and sparse classes dataset.

cs.CV↗

arXiv2018

Ultrafast hot carrier dynamics of ZrTe$_5$ from time-resolved optical reflectivity

We investigate the hot carrier dynamics of ZrTe$_5$ by ultrafast time-resolved optical reflectivity. Our results reveal a phonon-mediated across-gap recombination, consistent with its temperature-dependent gap nature as observed previously by photoemission. In addition, two distinct relaxations with a kink feature right after initial photoexcitation are well resolved, suggesting the complexity of electron thermalization process. Our findings indicate that correlated many-body effects play important role for the transient dynamics of ZrTe$_5$.

cond-mat.str-el↗

arXiv2018

Microscopy of hydrogen and hydrogen-vacancy defect structures on graphene devices

We have used scanning tunneling microscopy (STM) to investigate two types of hydrogen defect structures on monolayer graphene supported by hexagonal boron nitride (h-BN) in a gated field-effect transistor configuration. The first H-defect type is created by bombarding graphene with 1-keV ionized hydrogen and is identified as two hydrogen atoms bonded to a graphene vacancy via comparison of experimental data to first-principles calculations. The second type of H defect is identified as dimerized hydrogen and is created by depositing atomic hydrogen having only thermal energy onto a graphene surface. Scanning tunneling spectroscopy (STS) measurements reveal that hydrogen dimers formed in this way open a new elastic channel in the tunneling conductance between an STM tip and graphene.

cond-mat.mes-hall↗

arXiv2018

Quantum Error Correction with the Toric-GKP Code

We examine the performance of the single-mode GKP code and its concatenation with the toric code for a noise model of Gaussian shifts, or displacement errors. We show how one can optimize the tracking of errors in repeated noisy error correction for the GKP code. We do this by examining the maximum-likelihood problem for this setting and its mapping onto a 1D Euclidean path-integral modeling a particle in a random cosine potential. We demonstrate the efficiency of a minimum-energy decoding strategy as a proxy for the path integral evaluation. In the second part of this paper, we analyze and numerically assess the concatenation of the GKP code with the toric code. When toric code measurements and GKP error correction measurements are perfect, we find that by using GKP error information the toric code threshold improves from $10\%$ to $14\%$. When only the GKP error correction measurements are perfect we observe a threshold at $6\%$. In the more realistic setting when all error information is noisy, we show how to represent the maximum likelihood decoding problem for the toric-GKP code as a 3D compact QED model in the presence of a quenched random gauge field, an extension of the random-plaquette gauge model for the toric code. We present a new decoder for this problem which shows the existence of a noise threshold at shift-error standard deviation $σ_0 \approx 0.243$ for toric code measurements, data errors and GKP ancilla errors. If the errors only come from having imperfect GKP states, this corresponds to states with just 4 photons or more. Our last result is a no-go result for linear oscillator codes, encoding oscillators into oscillators. For the Gaussian displacement error model, we prove that encoding corresponds to squeezing the shift errors. This shows that linear oscillator codes are useless for quantum information protection against Gaussian shift errors.

quant-ph↗

arXiv2018

Where-and-When to Look: Deep Siamese Attention Networks for Video-based Person Re-identification

Video-based person re-identification (re-id) is a central application in surveillance systems with significant concern in security. Matching persons across disjoint camera views in their video fragments is inherently challenging due to the large visual variations and uncontrolled frame rates. There are two steps crucial to person re-id, namely discriminative feature learning and metric learning. However, existing approaches consider the two steps independently, and they do not make full use of the temporal and spatial information in videos. In this paper, we propose a Siamese attention architecture that jointly learns spatiotemporal video representations and their similarity metrics. The network extracts local convolutional features from regions of each frame, and enhance their discriminative capability by focusing on distinct regions when measuring the similarity with another pedestrian video. The attention mechanism is embedded into spatial gated recurrent units to selectively propagate relevant features and memorize their spatial dependencies through the network. The model essentially learns which parts (\emph{where}) from which frames (\emph{when}) are relevant and distinctive for matching persons and attaches higher importance therein. The proposed Siamese model is end-to-end trainable to jointly learn comparable hidden representations for paired pedestrian videos and their similarity value. Extensive experiments on three benchmark datasets show the effectiveness of each component of the proposed deep network while outperforming state-of-the-art methods.

cs.CV↗

arXiv2018

Joint Unsupervised Learning of Optical Flow and Depth by Watching Stereo Videos

Learning depth and optical flow via deep neural networks by watching videos has made significant progress recently. In this paper, we jointly solve the two tasks by exploiting the underlying geometric rules within stereo videos. Specifically, given two consecutive stereo image pairs from a video, we first estimate depth, camera ego-motion and optical flow from three neural networks. Then the whole scene is decomposed into moving foreground and static background by compar- ing the estimated optical flow and rigid flow derived from the depth and ego-motion. We propose a novel consistency loss to let the optical flow learn from the more accurate rigid flow in static regions. We also design a rigid alignment module which helps refine ego-motion estimation by using the estimated depth and optical flow. Experiments on the KITTI dataset show that our results significantly outperform other state-of- the-art algorithms. Source codes can be found at https: //github.com/baidu-research/UnDepthflow

cs.CV↗

arXiv2018

The Three Hundred project: a large catalogue of theoretically modelled galaxy clusters for cosmological and astrophysical applications

We introduce the THE THREE HUNDRED project, an endeavour to model 324 large galaxy clusters with full-physics hydrodynamical re-simulations. Here we present the data set and study the differences to observations for fundamental galaxy cluster properties and scaling relations. We find that the modelled galaxy clusters are generally in reasonable agreement with observations with respect to baryonic fractions and gas scaling relations at redshift z = 0. However, there are still some (model-dependent) differences, such as central galaxies being too massive, and galaxy colours (g - r) being bluer (about 0.2 dex lower at the peak position) than in observations. The agreement in gas scaling relations down to 10^{13} h^{-1} M_{\odot} between the simulations indicates that particulars of the sub-grid modelling of the baryonic physics only has a weak influence on these relations. We also include - where appropriate - a comparison to three semi-analytical galaxy formation models as applied to the same underlying dark-matter-only simulation. All simulations and derived data products are publicly available.

astro-ph.GA↗

arXiv2018

Model updating using sum of squares (SOS) optimization to minimize modal dynamic residuals

This research studies finite element (FE) model updating through sum of squares (SOS) optimization to minimize modal dynamic residuals. In the past few decades, many FE model updating algorithms have been studied to improve the similitude between a numerical model and the as-built structure. FE model updating usually requires solving nonconvex optimization problems, while most off-the-shelf optimization solvers can only find local optima. To improve the model updating performance, this paper proposes the SOS global optimization method for minimizing modal dynamic residuals of the generalized eigenvalue equations in structural dynamics. The proposed method is validated through both numerical simulation and experimental study of a four-story shear frame structure.

math.OC↗

arXiv2018

Video Summarization Using Fully Convolutional Sequence Networks

This paper addresses the problem of video summarization. Given an input video, the goal is to select a subset of the frames to create a summary video that optimally captures the important information of the input video. With the large amount of videos available online, video summarization provides a useful tool that assists video search, retrieval, browsing, etc. In this paper, we formulate video summarization as a sequence labeling problem. Unlike existing approaches that use recurrent models, we propose fully convolutional sequence models to solve video summarization. We firstly establish a novel connection between semantic segmentation and video summarization, and then adapt popular semantic segmentation networks for video summarization. Extensive experiments and analysis on two benchmark datasets demonstrate the effectiveness of our models.

cs.CV↗

arXiv2018

Fully Point-wise Convolutional Neural Network for Modeling Statistical Regularities in Natural Images

Modeling statistical regularity plays an essential role in ill-posed image processing problems. Recently, deep learning based methods have been presented to implicitly learn statistical representation of pixel distributions in natural images and leverage it as a constraint to facilitate subsequent tasks, such as color constancy and image dehazing. However, the existing CNN architecture is prone to variability and diversity of pixel intensity within and between local regions, which may result in inaccurate statistical representation. To address this problem, this paper presents a novel fully point-wise CNN architecture for modeling statistical regularities in natural images. Specifically, we propose to randomly shuffle the pixels in the origin images and leverage the shuffled image as input to make CNN more concerned with the statistical properties. Moreover, since the pixels in the shuffled image are independent identically distributed, we can replace all the large convolution kernels in CNN with point-wise ($1*1$) convolution kernels while maintaining the representation ability. Experimental results on two applications: color constancy and image dehazing, demonstrate the superiority of our proposed network over the existing architectures, i.e., using 1/10$\sim$1/100 network parameters and computational cost while achieving comparable performance.

cs.CV↗

arXiv2018

A Benchmark Test of Boson Sampling on Tianhe-2 Supercomputer

Boson sampling, thought to be intractable classically, can be solved by a quantum machine composed of merely generation, linear evolution and detection of single photons. Such an analog quantum computer for this specific problem provides a shortcut to boost the absolute computing power of quantum computers to beat classical ones. However, the capacity bound of classical computers for simulating boson sampling has not yet been identified. Here we simulate boson sampling on the Tianhe-2 supercomputer which occupied the first place in the world ranking six times from 2013 to 2016. We computed the permanent of the largest matrix using up to 312,000 CPU cores of Tianhe-2, and inferred from the current most efficient permanent-computing algorithms that an upper bound on the performance of Tianhe-2 is one 50-photon sample per ~100 min. In addition, we found a precision issue with one of two permanent-computing algorithms.

quant-ph↗

arXiv2018

Every Pixel Counts: Unsupervised Geometry Learning with Holistic 3D Motion Understanding

Learning to estimate 3D geometry in a single image by watching unlabeled videos via deep convolutional network has made significant process recently. Current state-of-the-art (SOTA) methods, are based on the learning framework of rigid structure-from-motion, where only 3D camera ego motion is modeled for geometry estimation.However, moving objects also exist in many videos, e.g. moving cars in a street scene. In this paper, we tackle such motion by additionally incorporating per-pixel 3D object motion into the learning framework, which provides holistic 3D scene flow understanding and helps single image geometry estimation. Specifically, given two consecutive frames from a video, we adopt a motion network to predict their relative 3D camera pose and a segmentation mask distinguishing moving objects and rigid background. An optical flow network is used to estimate dense 2D per-pixel correspondence. A single image depth network predicts depth maps for both images. The four types of information, i.e. 2D flow, camera pose, segment mask and depth maps, are integrated into a differentiable holistic 3D motion parser (HMP), where per-pixel 3D motion for rigid background and moving objects are recovered. We design various losses w.r.t. the two types of 3D motions for training the depth and motion networks, yielding further error reduction for estimated geometry. Finally, in order to solve the 3D motion confusion from monocular videos, we combine stereo images into joint training. Experiments on KITTI 2015 dataset show that our estimated geometry, 3D motion and moving object masks, not only are constrained to be consistent, but also significantly outperforms other SOTA algorithms, demonstrating the benefits of our approach.

cs.CV↗

arXiv2018

Real-time fMRI-based Brain Computer Interface: A Review

In recent years, the rapid development of neuroimaging technology has been providing many powerful tools for cognitive neuroscience research. Among them, the functional magnetic resonance imaging (fMRI), which has high spatial resolution, acceptable temporal resolution, simple calibration, and short preparation time, has been widely used in brain research. Compared with the electroencephalogram (EEG), real-time fMRI-based brain computer interface (rtfMRI-BCI) not only can perform decoding analysis across the whole brain to control external devices, but also allows a subject to voluntarily self-regulate specific brain regions. This paper reviews the basic architecture of rtfMRI-BCI, the emerging machine learning based data analysis approaches (also known as multi-voxel pattern analysis), and the applications and recent advances of rtfMRI-BCI.

cs.HC↗

arXiv2018

Manifold: A Model-Agnostic Framework for Interpretation and Diagnosis of Machine Learning Models

Interpretation and diagnosis of machine learning models have gained renewed interest in recent years with breakthroughs in new approaches. We present Manifold, a framework that utilizes visual analysis techniques to support interpretation, debugging, and comparison of machine learning models in a more transparent and interactive manner. Conventional techniques usually focus on visualizing the internal logic of a specific model type (i.e., deep neural networks), lacking the ability to extend to a more complex scenario where different model types are integrated. To this end, Manifold is designed as a generic framework that does not rely on or access the internal logic of the model and solely observes the input (i.e., instances or features) and the output (i.e., the predicted result and probability distribution). We describe the workflow of Manifold as an iterative process consisting of three major phases that are commonly involved in the model development and diagnosis process: inspection (hypothesis), explanation (reasoning), and refinement (verification). The visual components supporting these tasks include a scatterplot-based visual summary that overviews the models' outcome and a customizable tabular view that reveals feature discrimination. We demonstrate current applications of the framework on the classification and regression tasks and discuss other potential machine learning use scenarios where Manifold can be applied.

cs.LG↗

arXiv2018

A Wavefunction Microscope for Ultracold Atoms

Quantum simulations with ultracold atoms typically create atomic wavefunctions with structures at optical length scales, where direct imaging suffers from the diffraction limit. In analogy to advances in optical microscopy for biological applications, we use a non-linear atomic response to surpass the diffraction limit. Exploiting quantum interference, we demonstrate imaging with super-resolution of $λ/100$ and excellent temporal resolution of 500 ns. We characterize our microscope's performance by measuring the ensemble averaged wavefunction of atoms within the unit cells of an optical lattice, and observe the dynamics of atoms excited into periodic motion. This approach can be readily applied to image any atomic or molecular system, as long as it hosts a three-level system.

cond-mat.quant-gas↗