arXiv subjects

Bo Zhang

Publications and source records attributed to Bo Zhang.

At least 793 records · Page 44Linked to original sources

arXiv2017

Recovering scattering obstacles by multi-frequency phaseless far-field data

It is well known that the modulus of the far-field pattern (or phaseless far-field pattern) is invariant under translations of the scattering obstacle if only one plane wave is used as the incident field, so the shape but not the location of the obstacle can be recovered from the phaseless far-field data. In this paper, it is proved that the translation invariance property of the phaseless far-field pattern can be broken if superpositions of two plane waves are used as the incident fields for all wave numbers in a finite interval. Based on this, a recursive Newton-type iteration algorithm in frequencies is then developed to recover both the location and the shape of the obstacle simultaneously from multi-frequency phaseless far-field data. Numerical examples are also carried out to illustrate the validity of the approach and the effectiveness of the inversion algorithm.

math.NA↗

arXiv2017

Dust cluster spin in complex (dusty) plasmas

The spontaneous rotation of small dust clusters confined inside a cubical glass box in the sheath of a complex plasma was observed in experiment. Due to strong coupling between the dust particles, these clusters behave like a rigid-body where cluster rotation is contingent upon their configuration and symmetry. By evaluating the effects of distinct contributing forces, it is postulated that the rotation observed is driven by the net torque exerted on the cluster by the ion wake force. The configuration and symmetry of a cluster determines whether the net torque induced by the ion wake force is nonzero, in turn leading to cluster rotation. A COPTIC (Cartesian mesh, oblique boundary, particles and thermals in cell) simulation is employed to obtain the ion wake potential providing a theoretical model of cluster rotation which includes both the ion wake force and neutral drag and predicts rotation rates and direction in agreement with experimental results. These results are then used to diagnose the ion flow within the box.

physics.plasm-ph↗

arXiv2017

Multiscale Simulation of Surface Nanostructure Effect on Bubble Nucleation

Effects of nanostructured defects of copper solid surface on the bubble growth in liquid argon have been investigated through a hybrid atomistic-continuum method. The same solid surfaces with five different nanostructures, namely, wedge defect, deep rectangular defect (R-I), shallow rectangular defect (R-II), small rectangular defect (R-III) and no defect, have been modeled at molecular level. The liquid argon is placed on top of the hot solid copper with superheat of 30 K after equilibration is achieved with CFD-MD coupled simulation. Phase change of argon on five nanostructures has been observed and analyzed accordingly. The results showed that the solid surface with wedge defect tends to induce a nano-bubble relatively more easily than the others, and the larger the size of the defect is the easier the bubble generate.

physics.flu-dyn↗

arXiv2017

Universal Quantum Control in Zero-field Nuclear Magnetic Resonance

This paper describes a general method for manipulation of nuclear spins in zero magnetic field. In the absence of magnetic fields, the spins lose the individual information on chemical shifts and inequivalent spins can only be distinguished by nuclear gyromagnetic ratios and spin-spin couplings. For spin-1/2 nuclei with different gyromagnetic ratios (i.e., different species) in zero magnetic field, we describe the scheme to realize a set of universal quantum logic gates, e.g., arbitrary single-qubit gates and two-qubit controlled-NOT gate. This method allows for universal quantum control in systems which might provide promising applications in materials science, chemistry, biology,quantum information processing and fundamental physics.

quant-ph↗

arXiv2017

CNN Feature boosted SeqSLAM for Real-Time Loop Closure Detection

Loop closure detection (LCD) is an indispensable part of simultaneous localization and mapping systems (SLAM); it enables robots to produce a consistent map by recognizing previously visited places. When robots operate over extended periods, robustness to viewpoint and condition changes as well as satisfactory real-time performance become essential requirements for a practical LCD system. This paper presents an approach to directly utilize the outputs at the intermediate layer of a pre-trained convolutional neural network (CNN) as image descriptors. The matching location is determined by matching the image sequences through a method called SeqCNNSLAM. The utility of SeqCNNSLAM is comprehensively evaluated in terms of viewpoint and condition invariance. Experiments show that SeqCNNSLAM outperforms state-of-the-art LCD systems, such as SeqSLAM and Change Removal, in most cases. To allow for the real-time performance of SeqCNNSLAM, an acceleration method, A-SeqCNNSLAM, is established. This method exploits the location relationship between the matching images of adjacent images to reduce the matching range of the current image. Results demonstrate that acceleration of 4-6 is achieved with minimal accuracy degradation, and the method's runtime satisfies the real-time demand. To extend the applicability of A-SeqCNNSLAM to new environments, a method called O-SeqCNNSLAM is established for the online adjustment of the parameters of A-SeqCNNSLAM.

cs.RO↗

arXiv2017

Angstrom-resolution single-molecule fluorescence resonance energy transfer reveals mechanisms of DNA helicases

Single-molecule FRET is widely used to study helicases by detecting distance changes between a fluorescent donor and an acceptor anchored to overhangs of a forked DNA duplex. However, it has lacked single-base pair (1-bp) resolution required for revealing stepping dynamics in unwinding because FRET signals are usually blurred by thermal fluctuations of the overhangs. We designed a nanotensioner in which a short DNA is bent to exert a force on the overhangs, just as in optical/magnetic tweezers. The strategy improved the resolution of FRET to 0.5 bp, high enough to uncover the differences in DNA unwinding by yeast Pif1 and E. coli RecQ whose unwinding behaviors cannot be differentiated by currently practiced methods. We found that Pif1 exhibits 1-bp-stepping kinetics, while RecQ breaks 1 bp at a time but questers the nascent nucleotides and releases them randomly. The high-resolution data allowed us to propose a three-parameter model to quantitatively interpret the apparently different unwinding behaviors of the two helicases which belong to two superfamilies.

physics.bio-ph↗

arXiv2017

Improving Interpretability of Deep Neural Networks with Semantic Information

Interpretability of deep neural networks (DNNs) is essential since it enables users to understand the overall strengths and weaknesses of the models, conveys an understanding of how the models will behave in the future, and how to diagnose and correct potential problems. However, it is challenging to reason about what a DNN actually does due to its opaque or black-box nature. To address this issue, we propose a novel technique to improve the interpretability of DNNs by leveraging the rich semantic information embedded in human descriptions. By concentrating on the video captioning task, we first extract a set of semantically meaningful topics from the human descriptions that cover a wide range of visual concepts, and integrate them into the model with an interpretive loss. We then propose a prediction difference maximization algorithm to interpret the learned features of each neuron. Experimental results demonstrate its effectiveness in video captioning using the interpretable features, which can also be transferred to video action recognition. By clearly understanding the learned features, users can easily revise false predictions via a human-in-the-loop procedure.

cs.CV↗

arXiv2017

Discriminatively Boosted Image Clustering with Fully Convolutional Auto-Encoders

Traditional image clustering methods take a two-step approach, feature learning and clustering, sequentially. However, recent research results demonstrated that combining the separated phases in a unified framework and training them jointly can achieve a better performance. In this paper, we first introduce fully convolutional auto-encoders for image feature learning and then propose a unified clustering framework to learn image representations and cluster centers jointly based on a fully convolutional auto-encoder and soft $k$-means scores. At initial stages of the learning procedure, the representations extracted from the auto-encoder may not be very discriminative for latter clustering. We address this issue by adopting a boosted discriminative distribution, where high score assignments are highlighted and low score ones are de-emphasized. With the gradually boosted discrimination, clustering assignment scores are discriminated and cluster purities are enlarged. Experiments on several vision benchmark datasets show that our methods can achieve a state-of-the-art performance.

cs.CV↗

arXiv2017

Big Learning with Bayesian Methods

Explosive growth in data and availability of cheap computing resources have sparked increasing interest in Big learning, an emerging subfield that studies scalable machine learning algorithms, systems, and applications with Big Data. Bayesian methods represent one important class of statistic methods for machine learning, with substantial recent developments on adaptive, flexible and scalable Bayesian learning. This article provides a survey of the recent advances in Big learning with Bayesian methods, termed Big Bayesian Learning, including nonparametric Bayesian methods for adaptively inferring model complexity, regularized Bayesian inference for improving the flexibility via posterior regularization, and scalable algorithms and systems based on stochastic subsampling and distributed computing for dealing with large-scale applications.

cs.LG↗

arXiv2017

Dynamic self-assembly of charged colloidal strings and walls in simple fluid flows

Colloidal particles can self-assemble into various ordered structures in fluid flows that have potential applications in biomedicine, materials synthesis and encryption. These dynamic processes are also of fundamental interest for probing the general principles of self-assembly in non-equilibrium conditions. Here, we report a simple microfluidic experiment, where charged colloidal particles self-assemble into flow-aligned 1D strings with regular particle spacing near a solid boundary. Using high-speed confocal microscopy, we systematically investigate the influence of flow rates, electrostatics and particle polydispersity on the observed string structures. By studying the detailed dynamics of stable flow-driven particle pairs, we quantitatively characterize interparticle interactions. Based on the results, we construct a simple model that explains the intriguing non-equilibrium self-assembly process. Our study shows that the colloidal strings arise from a delicate balance between attractive hydrodynamic coupling and repulsive electrostatic interaction between particles. Finally, we demonstrate that, with the assistance of transverse electric fields, a similar mechanism also leads to the formation of 2D colloidal walls.

cond-mat.soft↗

arXiv2017

Estimation of Dynamic Mixed Double Factors Model in High Dimensional Panel Data

The purpose of this article is to develop the dimension reduction techniques in panel data analysis when the number of individuals and indicators is large. We use Principal Component Analysis (PCA) method to represent large number of indicators by minority common factors in the factor models. We propose the Dynamic Mixed Double Factor Model (DMDFM for short) to re ect cross section and time series correlation with interactive factor structure. DMDFM not only reduce the dimension of indicators but also consider the time series and cross section mixed effect. Different from other models, mixed factor model have two styles of common factors. The regressors factors re flect common trend and reduce the dimension, error components factors re ect difference and weak correlation of individuals. The results of Monte Carlo simulation show that Generalized Method of Moments (GMM) estimators have good unbiasedness and consistency. Simulation also shows that the DMDFM can improve prediction power of the models effectively.

stat.ME↗

arXiv2017

Direct and inverse scattering problem by an unbounded rough interface with buried obstacles

In this paper, we consider the direct and inverse problem of scattering of time-harmonic waves by an unbounded rough interface with a buried impenetrable obstacle. We first study the well-posedness of the direct problem with a local source by the variational method; the well-posedness result is then extended to scattering problems induced by point source waves (PSWs) and hyper-singular point source waves (HSPSWs). For PSW or HSPSW incident waves, the induced total field admits a uniformly bounded estimate in any compact subset far from the source position. Moreover, we show that the scattered field due to HSPSWs can be approximated by the scattered fields due to PSWs. With these properties and a novel reciprocity relation of the total field, we prove that the rough surface and the buried obstacle can be uniquely determined by the scattered near-field data measured only on a line segment above the rough surface. The proof substantially relies upon constructing a well-posed interior transmission problem for the Helmholtz equation.

math.AP↗

arXiv2016

Max-Margin Deep Generative Models for (Semi-)Supervised Learning

Deep generative models (DGMs) are effective on learning multilayered representations of complex data and performing inference of input data by exploring the generative ability. However, it is relatively insufficient to empower the discriminative ability of DGMs on making accurate predictions. This paper presents max-margin deep generative models (mmDGMs) and a class-conditional variant (mmDCGMs), which explore the strongly discriminative principle of max-margin learning to improve the predictive performance of DGMs in both supervised and semi-supervised learning, while retaining the generative capability. In semi-supervised learning, we use the predictions of a max-margin classifier as the missing labels instead of performing full posterior inference for efficiency; we also introduce additional max-margin and label-balance regularization terms of unlabeled data for effectiveness. We develop an efficient doubly stochastic subgradient algorithm for the piecewise linear objectives in different settings. Empirical results on various datasets demonstrate that: (1) max-margin learning can significantly improve the prediction performance of DGMs and meanwhile retain the generative ability; (2) in supervised learning, mmDGMs are competitive to the best fully discriminative networks when employing convolutional neural networks as the generative and recognition models; and (3) in semi-supervised learning, mmDCGMs can perform efficient inference and achieve state-of-the-art classification results on several benchmarks.

cs.CV↗

arXiv2016

Effective Deterministic Initialization for $k$-Means-Like Methods via Local Density Peaks Searching

The $k$-means clustering algorithm is popular but has the following main drawbacks: 1) the number of clusters, $k$, needs to be provided by the user in advance, 2) it can easily reach local minima with randomly selected initial centers, 3) it is sensitive to outliers, and 4) it can only deal with well separated hyperspherical clusters. In this paper, we propose a Local Density Peaks Searching (LDPS) initialization framework to address these issues. The LDPS framework includes two basic components: one of them is the local density that characterizes the density distribution of a data set, and the other is the local distinctiveness index (LDI) which we introduce to characterize how distinctive a data point is compared with its neighbors. Based on these two components, we search for the local density peaks which are characterized with high local densities and high LDIs to deal with 1) and 2). Moreover, we detect outliers characterized with low local densities but high LDIs, and exclude them out before clustering begins. Finally, we apply the LDPS initialization framework to $k$-medoids, which is a variant of $k$-means and chooses data samples as centers, with diverse similarity measures other than the Euclidean distance to fix the last drawback of $k$-means. Combining the LDPS initialization framework with $k$-means and $k$-medoids, we obtain two novel clustering methods called LDPS-means and LDPS-medoids, respectively. Experiments on synthetic data sets verify the effectiveness of the proposed methods, especially when the ground truth of the cluster number $k$ is large. Further, experiments on several real world data sets, Handwritten Pendigits, Coil-20, Coil-100 and Olivetti Face Database, illustrate that our methods give a superior performance than the analogous approaches on both estimating $k$ and unsupervised object categorization.

cs.LG↗

arXiv2016

Molecular Lines of 13 Galactic Infrared Bubble Regions

We investigated the physical properties of molecular clouds and star formation processes around infrared bubbles which are essentially expanding HII regions. We performed observations of 13 galactic infrared bubble fields containing 18 bubbles. Five molecular lines, 12CO (J=1-0), 13CO (J=1-0), C18O(J=1-0), HCN (J=1-0), and HCO+ (J=1-0), were observed, and several publicly available surveys, GLIMPSE, MIPSGAL, ATLASGAL, BGPS, VGPS, MAGPIS, and NVSS, were used for comparison. We find that these bubbles are generally connected with molecular clouds, most of which are giant. Several bubble regions display velocity gradients and broad shifted profiles, which could be due to the expansion of bubbles. The masses of molecular clouds within bubbles range from 100 to 19,000 solar mass, and their dynamic ages are about 0.3-3.7 Myr, which takes into account the internal turbulence pressure of surrounding molecular clouds. Clumps are found in the vicinity of all 18 bubbles, and molecular clouds near four of these bubbles with larger angular sizes show shell-like morphologies, indicating that either collect-and-collapse or radiation-driven implosion processes may have occurred. Due to the contamination of adjacent molecular clouds, only six bubble regions are appropriate to search for outflows, and we find that four of them have outflow activities. Three bubbles display ultra-compact HII regions at their borders, and one of them is probably responsible for its outflow. In total, only six bubbles show star formation activities in the vicinity, and we suggest that star formation processes might have been triggered.

astro-ph.GA↗

arXiv2016

Imaging of locally rough surfaces from intensity-only far-field or near-field data

This paper is concerned with a nonlinear imaging problem, which aims to reconstruct a locally perturbed, perfectly reflecting, infinite plane from intensity-only (or phaseless) far-field or near-field data. A recursive Newton iteration algorithm in frequencies is developed to reconstruct the locally rough surface from multi-frequency intensity-only far-field or near-field data, where the fast integral equation solver developed in [39] is used to solve the direct scattering problem in each iteration. For the case with far-field data, a main feature of our work is that the incident field is taken as a superposition of two plane waves with different directions rather than one plane wave, so the location and shape of the local perturbation of the infinite plane can be reconstructed simultaneously from intensity-only far-field data with multiple wave numbers. This is different from previous work on inverse scattering from phaseless far-field data, where only the shape reconstruction was considered due to the translation invariance property of the phaseless far-field pattern corresponding to one plane wave as the incident field. Finally, numerical examples are carried out to demonstrate that our reconstruction algorithm is stable and accurate even for the case of multiple-scale profiles.

math.NA↗

arXiv2016

Fast Sampling for Bayesian Max-Margin Models

Bayesian max-margin models have shown superiority in various practical applications, such as text categorization, collaborative prediction, social network link prediction and crowdsourcing, and they conjoin the flexibility of Bayesian modeling and predictive strengths of max-margin learning. However, Monte Carlo sampling for these models still remains challenging, especially for applications that involve large-scale datasets. In this paper, we present the stochastic subgradient Hamiltonian Monte Carlo (HMC) methods, which are easy to implement and computationally efficient. We show the approximate detailed balance property of subgradient HMC which reveals a natural and validated generalization of the ordinary HMC. Furthermore, we investigate the variants that use stochastic subsampling and thermostats for better scalability and mixing. Using stochastic subgradient Markov Chain Monte Carlo (MCMC), we efficiently solve the posterior inference task of various Bayesian max-margin models and extensive experimental results demonstrate the effectiveness of our approach.

stat.ML↗

arXiv2016

The local spiral structure of the Milky Way

The nature of the spiral structure of the Milky Way has long been debated. Only in the last decade have astronomers been able to accurately measure distances to a substantial number of high-mass star-forming regions, the classic tracers of spiral structure in galaxies. We report distance measurements at radio wavelengths using the Very Long Baseline Array for eight regions of massive star formation near the Local spiral arm of the Milky Way. Combined with previous measurements, these observations reveal that the Local Arm is larger than previously thought, and both its pitch angle and star formation rate are comparable to those of the Galaxy's major spiral arms, such as Sagittarius and Perseus. Toward the constellation Cygnus, sources in the Local Arm extend for a great distance along our line of sight and roughly along the solar orbit. Because of this orientation, these sources cluster both on the sky and in velocity to form the complex and long enigmatic Cygnus X region. We also identify a spur that branches between the Local and Sagittarius spiral arms.

astro-ph.GA↗