Searcharxiv⌕ Search

arXiv subjects

Jian Zhang

Publications and source records attributed to Jian Zhang.

At least 685 records · Page 38Linked to original sources
arXiv2017

Kill Two Birds with One Stone: Weakly-Supervised Neural Network for Image Annotation and Tag Refinement

The number of social images has exploded by the wide adoption of social networks, and people like to share their comments about them. These comments can be a description of the image, or some objects, attributes, scenes in it, which are normally used as the user-provided tags. However, it is well-known that user-provided tags are incomplete and imprecise to some extent. Directly using them can damage the performance of related applications, such as the image annotation and retrieval. In this paper, we propose to learn an image annotation model and refine the user-provided tags simultaneously in a weakly-supervised manner. The deep neural network is utilized as the image feature learning and backbone annotation model, while visual consistency, semantic dependency, and user-error sparsity are introduced as the constraints at the batch level to alleviate the tag noise. Therefore, our model is highly flexible and stable to handle large-scale image sets. Experimental results on two benchmark datasets indicate that our proposed model achieves the best performance compared to the state-of-the-art methods.

cs.CV↗
arXiv2017

Atomistic simulation of the Coupled adsorption and unfolding of protein GB1 on the polystyrenes nanoparticle surface

Protein adsorption/desorption upon nanoparticle surfaces is an important process to understand for developing new nanotechnology involving biomaterials, while atomistic picture of the process and its coupling with protein conformational change is lacking. Here we report our study on the adsorption of protein GB1 upon a polystyrene nanoparticle surface using atomistic molecular dynamic simulations. Enabled by metadynamics, we explored the relevant phase space and identified three protein states; each protein state involved both the adsorbed and desorbed states. We also studied the change of secondary and tertiary structures of GB1 during adsorption, and the dominant interactions between protein and surface in different adsorbing stages. From the simulation results we obtained a scenario that is more rational and complete than the conventional one. We believe the new scenario is more appropriate as a theoretical model in understanding and explaining experimental signals. Introduction

physics.bio-ph↗
arXiv2017

SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval

Hashing methods have been widely used for efficient similarity retrieval on large scale image database. Traditional hashing methods learn hash functions to generate binary codes from hand-crafted features, which achieve limited accuracy since the hand-crafted features cannot optimally represent the image content and preserve the semantic similarity. Recently, several deep hashing methods have shown better performance because the deep architectures generate more discriminative feature representations. However, these deep hashing methods are mainly designed for supervised scenarios, which only exploit the semantic similarity information, but ignore the underlying data structures. In this paper, we propose the semi-supervised deep hashing (SSDH) approach, to perform more effective hash function learning by simultaneously preserving semantic similarity and underlying data structures. The main contributions are as follows: (1) We propose a semi-supervised loss to jointly minimize the empirical error on labeled data, as well as the embedding error on both labeled and unlabeled data, which can preserve the semantic similarity and capture the meaningful neighbors on the underlying data structures for effective hashing. (2) A semi-supervised deep hashing network is designed to extensively exploit both labeled and unlabeled data, in which we propose an online graph construction method to benefit from the evolving deep features during training to better capture semantic neighbors. To the best of our knowledge, the proposed deep network is the first deep hashing method that can perform hash code learning and feature learning simultaneously in a semi-supervised fashion. Experimental results on 5 widely-used datasets show that our proposed approach outperforms the state-of-the-art hashing methods.

cs.CV↗
arXiv2017

In situ high resolution real-time quantum efficiency imaging for photocathodes

Aspects of the preparation process and performance degradation are two major problems of photocathodes. The lack of a means for dynamic quantum efficiency measurements results in the inability to observe the inhomogeneity of the cathode surface by fine structural analysis and in real time.Here we present a simple and scalable technique for in situ real-time quantum efficiency diagnosis. An incoherent light source provides uniform illumination on the cathode surface, and solenoid magnets are used as lens for focusing and imaging the emitted electron beam on a downstream scintillator screen, which converts the quantum efficiency information into fluorescence intensity distribution. The microscopic discontinuity and the dynamic changes of the quantum efficiency of a gallium arsenide photocathode are observed at a resolution of a few microns. An unexpected uneven decrease of the quantum efficiency is also recorded. The work demonstrates a new observation method for photoemission materials research.

physics.app-ph↗
arXiv2017

Depth dependent decoherence caused by surface and external spins for NV centers in diamond

By efficient nanoscale plasma etching, the nitrogen-vacancy (NV) centers in diamond were brought to the sample surface step by step successfully. At each depth, we used the relative ratios of spin coherence times before and after applying external spins on the surface to present the decoherence, and investigated the relationships between depth and ratios. The values of relative ratios declined and then rised with the decreasing depth, which was attributed to the decoherence influenced by external spins, surface spins, discrete surface spin effects and electric field noise. Moreover, our work revealed a characteristic depth at which the NV center would experience relatively the strongest decoherence caused by external spins in consideration of inevitable surface spins. And the characteristic depth was found depending on the adjacent environments of NV centers and the density of surface spins.

cond-mat.mes-hall↗
arXiv2017

Ultra-low Power Nanoelectromechanical Memory Based on Location-controllable Nanogap System

Nanogap engineering of low-dimensional nanomaterials, has received considerable interest in a variety of fields, ranging from molecular electronics to memories. Creating nanogaps at a certain position is of vital importance for the repeatable fabrication of the devices. In this work, we report a rational design of non-volatile memories based on sub-5 nm nanogaped single-walled carbon nanotubes (SWNTs) via the electromechanical motion. The nanogaps are readily realized by electroburning in a partially suspended SWNTs device with nanoscale region. The SWNT memory devices are applicable for both metallic and semiconducting SWNTs, resolving the challenge of separation of semiconducting SWNTs from metallic ones. Meanwhile, the memory devices exhibit excellent performance: ultra-low writing energy (4.1*10-19 J per bit), ON/OFF ratio of 105, stable switching ON operations and over 30 hours retention time in ambient conditions, which are of great potential for applications.

cond-mat.mes-hall↗
arXiv2017

Target Directed Event Sequence Generation for Android Applications

Testing is a commonly used approach to ensure the quality of software, of which model-based testing is a hot topic to test GUI programs such as Android applications (apps). Existing approaches mainly either dynamically construct a model that only contains the GUI information, or build a model in the view of code that may fail to describe the changes of GUI widgets during runtime. Besides, most of these models do not support back stack that is a particular mechanism of Android. Therefore, this paper proposes a model LATTE that is constructed dynamically with consideration of the view information in the widgets as well as the back stack, to describe the transition between GUI widgets. We also propose a label set to link the elements of the LATTE model to program snippets. The user can define a subset of the label set as a target for the testing requirements that need to cover some specific parts of the code. To avoid the state explosion problem during model construction, we introduce a definition "state similarity" to balance the model accuracy and analysis cost. Based on this model, a target directed test generation method is presented to generate event sequences to effectively cover the target. The experiments on several real-world apps indicate that the generated test cases based on LATTE can reach a high coverage, and with the model we can generate the event sequences to cover a given target with short event sequences.

cs.SE↗
arXiv2017

Deep Learning at 15PF: Supervised and Semi-Supervised Classification for Scientific Data

This paper presents the first, 15-PetaFLOP Deep Learning system for solving scientific pattern classification problems on contemporary HPC architectures. We develop supervised convolutional architectures for discriminating signals in high-energy physics data as well as semi-supervised architectures for localizing and classifying extreme weather in climate data. Our Intelcaffe-based implementation obtains $\sim$2TFLOP/s on a single Cori Phase-II Xeon-Phi node. We use a hybrid strategy employing synchronous node-groups, while using asynchronous communication across groups. We use this strategy to scale training of a single model to $\sim$9600 Xeon-Phi nodes; obtaining peak performance of 11.73-15.07 PFLOP/s and sustained performance of 11.41-13.27 PFLOP/s. At scale, our HEP architecture produces state-of-the-art classification accuracy on a dataset with 10M images, exceeding that achieved by selections on high-level physics-motivated features. Our semi-supervised architecture successfully extracts weather patterns in a 15TB climate dataset. Our results demonstrate that Deep Learning can be optimized and scaled effectively on many-core, HPC systems.

cs.PF↗
arXiv2017

Gelfand-Tsetlin modules of quantum gl_n$defined by admissible sets of relations

The purpose of this paper is to construct new families of irreducible Gelfand-Tsetlin modules for U_q(gl_n). These modules have arbitrary singularity and Gelfand-Tsetlin multiplicities bounded by 2. Most previously known irreducible modules had all Gelfand-Tsetlin multiplicities bounded by 1 \cite{FRZ1}, \cite{FRZ2}. In particular, our method works for q=1 providing new families of irreducible Gelfand-Tsetlin modules for gl_n. This generalizes the results of \cite{FGR3} and \cite{FRZ}.

math.RT↗
arXiv2017

Discovery of slow magnetic fluctuations and critical slowing down in the pseudogap phase of YBa$_2$Cu$_3$O$_y$

Evidence for intra-unit-cell (IUC) magnetic order in the pseudogap region of high-$T_c$ cuprates below a temperature $T^\ast$ is found in several studies, but NMR and $μ$SR experiments do not observe the expected static local magnetic fields. It has been noted, however, that such fields could be averaged by fluctuations. Our measurements of muon spin relaxation rates in single crystals of YBa$_2$Cu$_3$O$_y$ reveal magnetic fluctuations of the expected order of magnitude that exhibit critical slowing down at $T^\ast$. These results are strong evidence for fluctuating IUC magnetic order in the pseudogap phase.

cond-mat.supr-con↗
arXiv2017

A New Probabilistic Algorithm for Approximate Model Counting

Constrained counting is important in domains ranging from artificial intelligence to software analysis. There are already a few approaches for counting models over various types of constraints. Recently, hashing-based approaches achieve both theoretical guarantees and scalability, but still rely on solution enumeration. In this paper, a new probabilistic polynomial time approximate model counter is proposed, which is also a hashing-based universal framework, but with only satisfiability queries. A variant with a dynamic stopping criterion is also presented. Empirical evaluation over benchmarks on propositional logic formulas and SMT(BV) formulas shows that the approach is promising.

cs.AI↗
arXiv2017

Chemical reaction directed oriented attachment: from precursor particles to new substances

The oriented attachment (OA) of nanoparticles is an important mechanism for the synthesis of the crystals of inorganic functional materials, and the formation of natural minerals. For years it has been generally acknowledged that OA is a physical process, i.e., particle alignments and interface fusion via mass diffusion, not involving the formation of new substances. Hence, the obtained crystals maintain identical crystallographic structures and chemical constituents to those of the precursor particles. Here we report a chemical reaction directed OA growth, through which Y2(CO3)3.2H2O nanoparticles are converted to single-crystalline double-carbonates (e.g., NaY(CO3)2.6H2O). The dominant role of OA growth is supported by our first-principles calculations. Such a new OA mechanism enriches the aggregation-based crystal growth theory.

cond-mat.mtrl-sci↗
arXiv2017

Query-adaptive Image Retrieval by Deep Weighted Hashing

Hashing methods have attracted much attention for large scale image retrieval. Some deep hashing methods have achieved promising results by taking advantage of the strong representation power of deep networks recently. However, existing deep hashing methods treat all hash bits equally. On one hand, a large number of images share the same distance to a query image due to the discrete Hamming distance, which raises a critical issue of image retrieval where fine-grained rankings are very important. On the other hand, different hash bits actually contribute to the image retrieval differently, and treating them equally greatly affects the retrieval accuracy of image. To address the above two problems, we propose the query-adaptive deep weighted hashing (QaDWH) approach, which can perform fine-grained ranking for different queries by weighted Hamming distance. First, a novel deep hashing network is proposed to learn the hash codes and corresponding class-wise weights jointly, so that the learned weights can reflect the importance of different hash bits for different image classes. Second, a query-adaptive image retrieval method is proposed, which rapidly generates hash bit weights for different query images by fusing its semantic probability and the learned class-wise weights. Fine-grained image retrieval is then performed by the weighted Hamming distance, which can provide more accurate ranking than the traditional Hamming distance. Experiments on four widely used datasets show that the proposed approach outperforms eight state-of-the-art hashing methods.

cs.CV↗
arXiv2017

TruthDiscover: Resolving Object Conflicts on Massive Linked Data

Considerable effort has been made to increase the scale of Linked Data. However, because of the openness of the Semantic Web and the ease of extracting Linked Data from semi-structured sources (e.g., Wikipedia) and unstructured sources, many Linked Data sources often provide conflicting objects for a certain predicate of a real-world entity. Existing methods cannot be trivially extended to resolve conflicts in Linked Data because Linked Data has a scale-free property. In this demonstration, we present a novel system called TruthDiscover, to identify the truth in Linked Data with a scale-free property. First, TruthDiscover leverages the topological properties of the Source Belief Graph to estimate the priori beliefs of sources, which are utilized to smooth the trustworthiness of sources. Second, the Hidden Markov Random Field is utilized to model interdependencies among objects for estimating the trust values of objects accurately. TruthDiscover can visualize the process of resolving conflicts in Linked Data. Experiments results on four datasets show that TruthDiscover exhibits satisfactory accuracy when confronted with data having a scale-free property.

cs.DB↗
arXiv2017

Explicit construction of irreducible modules for U_q(gl_n)

We construct new families of U_q(gl_n)-modules by continuation from finite dimensional representations. Each such module is associated with a combinatorial object - admissible set of relations defined in \cite{FRZ}. More precisely, we prove that any admissible set of relations leads to a family of irreducible U_q(gl_n)-modules. Finite dimensional and generic modules are particular cases of this construction.

math.RT↗
arXiv2017

A hybrid structure gaseous detector for ion backflow suppression*

A new concept for a hybrid structure gaseous detector module with ion backflow suppression for the time projection chamber in a future circular collider is presented. It is a hybrid structure cascaded Gas Electron Multiplier (GEM) with a Micromegas detector. Both Micromegas and GEM have the capability to naturally reduce most of the ions produced in the amplification region. The GEM also acts as the preamplifer device and increases gas gain together with the Micromegas. Feasibility tests of the hybrid detector are performed using an ${}^{55}$Fe X-ray source. The energy resolution is better than 27% for 5.9 keV X-rays. It is demonstrated that a backflow ratio better than 0.2% can be reached in the hybrid readout structure at a gain of 5000.

physics.ins-det↗
arXiv2017

Exploiting Web Images for Dataset Construction: A Domain Robust Approach

Labelled image datasets have played a critical role in high-level image understanding. However, the process of manual labelling is both time-consuming and labor intensive. To reduce the cost of manual labelling, there has been increased research interest in automatically constructing image datasets by exploiting web images. Datasets constructed by existing methods tend to have a weak domain adaptation ability, which is known as the "dataset bias problem". To address this issue, we present a novel image dataset construction framework that can be generalized well to unseen target domains. Specifically, the given queries are first expanded by searching the Google Books Ngrams Corpus to obtain a rich semantic description, from which the visually non-salient and less relevant expansions are filtered out. By treating each selected expansion as a "bag" and the retrieved images as "instances", image selection can be formulated as a multi-instance learning problem with constrained positive bags. We propose to solve the employed problems by the cutting-plane and concave-convex procedure (CCCP) algorithm. By using this approach, images from different distributions can be kept while noisy images are filtered out. To verify the effectiveness of our proposed approach, we build an image dataset with 20 categories. Extensive experiments on image classification, cross-dataset generalization, diversity comparison and object detection demonstrate the domain robustness of our dataset.

cs.CV↗