Showing papers by "Tsinghua University published in 2015"

Scalable Person Re-identification: A Benchmark

Abstract: This paper contributes a new high quality dataset for person re-identification, named "Market-1501". Generally, current datasets: 1) are limited in scale, 2) consist of hand-drawn bboxes, which are unavailable under realistic settings, 3) have only one ground truth and one query image for each identity (close environment). To tackle these problems, the proposed Market-1501 dataset is featured in three aspects. First, it contains over 32,000 annotated bboxes, plus a distractor set of over 500K images, making it the largest person re-id dataset to date. Second, images in Market-1501 dataset are produced using the Deformable Part Model (DPM) as pedestrian detector. Third, our dataset is collected in an open system, where each identity has multiple images under each camera. As a minor contribution, inspired by recent advances in large-scale image search, this paper proposes an unsupervised Bag-of-Words descriptor. We view person re-identification as a special task of image search. In experiment, we show that the proposed descriptor yields competitive accuracy on VIPeR, CUHK03, and Market-1501 datasets, and is scalable on the large-scale 500k dataset.

Cleavage of GSDMD by inflammatory caspases determines pyroptotic cell death

Abstract: Inflammatory caspases (caspase-1, -4, -5 and -11) are critical for innate defences. Caspase-1 is activated by ligands of various canonical inflammasomes, and caspase-4, -5 and -11 directly recognize bacterial lipopolysaccharide, both of which trigger pyroptosis. Despite the crucial role in immunity and endotoxic shock, the mechanism for pyroptosis induction by inflammatory caspases is unknown. Here we identify gasdermin D (Gsdmd) by genome-wide clustered regularly interspaced palindromic repeat (CRISPR)-Cas9 nuclease screens of caspase-11- and caspase-1-mediated pyroptosis in mouse bone marrow macrophages. GSDMD-deficient cells resisted the induction of pyroptosis by cytosolic lipopolysaccharide and known canonical inflammasome ligands. Interleukin-1β release was also diminished in Gsdmd(-/-) cells, despite intact processing by caspase-1. Caspase-1 and caspase-4/5/11 specifically cleaved the linker between the amino-terminal gasdermin-N and carboxy-terminal gasdermin-C domains in GSDMD, which was required and sufficient for pyroptosis. The cleavage released the intramolecular inhibition on the gasdermin-N domain that showed intrinsic pyroptosis-inducing activity. Other gasdermin family members were not cleaved by inflammatory caspases but shared the autoinhibition; gain-of-function mutations in Gsdma3 that cause alopecia and skin defects disrupted the autoinhibition, allowing its gasdermin-N domain to trigger pyroptosis. These findings offer insight into inflammasome-mediated immunity/diseases and also change our understanding of pyroptosis and programmed necrosis.

Learning Transferable Features with Deep Adaptation Networks

Abstract: Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multi-kernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks.

Learning entity and relation embeddings for knowledge graph completion

Abstract: Knowledge graph completion aims to perform link prediction between entities. In this paper, we consider the approach of knowledge graph embeddings. Recently, models such as TransE and TransH build entity and relation embeddings by regarding a relation as translation from head entity to tail entity. We note that these models simply put both entities and relations within the same semantic space. In fact, an entity may have multiple aspects and various relations may focus on different aspects of entities, which makes a common space insufficient for modeling. In this paper, we propose TransR to build entity and relation embeddings in separate entity space and relation spaces. Afterwards, we learn embeddings by first projecting entities from entity space to corresponding relation space and then building translations between projected entities. In experiments, we evaluate our models on three tasks including link prediction, triple classification and relational fact extraction. Experimental results show significant and consistent improvements compared to state-of-the-art baselines including TransE and TransH. The source code of this paper can be obtained from https://github.com/mrlyk423/relation_extraction.

Non-orthogonal multiple access for 5G: solutions, challenges, opportunities, and future research trends

Abstract: The increasing demand of mobile Internet and the Internet of Things poses challenging requirements for 5G wireless communications, such as high spectral efficiency and massive connectivity. In this article, a promising technology, non-orthogonal multiple access (NOMA), is discussed, which can address some of these challenges for 5G. Different from conventional orthogonal multiple access technologies, NOMA can accommodate much more users via nonorthogonal resource allocation. We divide existing dominant NOMA schemes into two categories: power-domain multiplexing and code-domain multiplexing, and the corresponding schemes include power-domain NOMA, multiple access with low-density spreading, sparse code multiple access, multi-user shared access, pattern division multiple access, and so on. We discuss their principles, key features, and pros/cons, and then provide a comprehensive comparison of these solutions from the perspective of spectral efficiency, system performance, receiver complexity, and so on. In addition, challenges, opportunities, and future research trends for NOMA design are highlighted to provide some insight on the potential future work for researchers in this field. Finally, to leverage different multiple access schemes including both conventional OMA and new NOMA, we propose the concept of software defined multiple access (SoDeMA), which enables adaptive configuration of available multiple access schemes to support diverse services and applications in future 5G networks.

Experimental Discovery of Weyl Semimetal TaAs

Abstract: Weyl fermions possess exotic properties and can act like magnetic monopoles. Researchers show that TaAs is a Weyl semimetal, demonstrating for the first time that Weyl semimetals can be identified experimentally.

Combined Measurement of the Higgs Boson Mass in pp Collisions at √s=7 and 8 TeV with the ATLAS and CMS Experiments

Abstract: A measurement of the Higgs boson mass is presented based on the combined data samples of the ATLAS and CMS experiments at the CERN LHC in the H→γγ and H→ZZ→4l decay channels. The results are obtained from a simultaneous fit to the reconstructed invariant mass peaks in the two channels and for the two experiments. The measured masses from the individual channels and the two experiments are found to be consistent among themselves. The combined measured mass of the Higgs boson is mH=125.09±0.21 (stat)±0.11 (syst) GeV.

Review of recent progress in chemical stability of perovskite solar cells

Abstract: In recent years, the record efficiency of perovskite solar cells (PSCs) has been updated from 9.7% to 20.1%. However, there has been very little study of the issue of stability, which restricts the outdoor application of PSCs. The issues of the degradation of perovskite and the stability of PSC devices should be urgently addressed to achieve good reproducibility and long lifetimes for PSCs with high conversion efficiency. Without studies on stability, exciting achievements cannot be transferred from the laboratory to industry and outdoor applications. In order to improve their stability, a basic understanding of the degradation process of PSCs in different conditions should be acquired via thorough study. This review summarizes recent studies of the relationship of the chemical stability of PSCs with their environment (oxygen and moisture, UV light, solution process, temperature) and corresponding possible solutions.

Chromatin architecture reorganization during stem cell differentiation

Abstract: Higher-order chromatin structure is emerging as an important regulator of gene expression. Although dynamic chromatin structures have been identified in the genome, the full scope of chromatin dynamics during mammalian development and lineage specification remains to be determined. By mapping genome-wide chromatin interactions in human embryonic stem (ES) cells and four human ES-cell-derived lineages, we uncover extensive chromatin reorganization during lineage specification. We observe that although self-associating chromatin domains are stable during differentiation, chromatin interactions both within and between domains change in a striking manner, altering 36% of active and inactive chromosomal compartments throughout the genome. By integrating chromatin interaction maps with haplotype-resolved epigenome and transcriptome data sets, we find widespread allelic bias in gene expression correlated with allele-biased chromatin states of linked promoters and distal enhancers. Our results therefore provide a global view of chromatin dynamics and a resource for studying long-range control of gene expression in distinct human cell lineages.

Learning Transferable Features with Deep Adaptation Networks

Abstract: Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multikernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain adaptation benchmarks.

Characterization of the LIGO detectors during their sixth science run

Abstract: In 2009–2010, the Laser Interferometer Gravitational-Wave Observatory (LIGO) operated together with international partners Virgo and GEO600 as a network to search for gravitational waves (GWs) of astrophysical origin. The sensitivity of these detectors was limited by a combination of noise sources inherent to the instrumental design and its environment, often localized in time or frequency, that couple into the GW readout. Here we review the performance of the LIGO instruments during this epoch, the work done to characterize the detectors and their data, and the effect that transient and continuous noise artefacts have on the sensitivity of LIGO to a variety of astrophysical sources.

Epitaxial growth of two-dimensional stanene

Abstract: Following the first experimental realization of graphene, other ultrathin materials with unprecedented electronic properties have been explored, with particular attention given to the heavy group-IV elements Si, Ge and Sn. Two-dimensional buckled Si-based silicene has been recently realized by molecular beam epitaxy growth, whereas Ge-based germanene was obtained by molecular beam epitaxy and mechanical exfoliation. However, the synthesis of Sn-based stanene has proved challenging so far. Here, we report the successful fabrication of 2D stanene by molecular beam epitaxy, confirmed by atomic and electronic characterization using scanning tunnelling microscopy and angle-resolved photoemission spectroscopy, in combination with first-principles calculations. The synthesis of stanene and its derivatives will stimulate further experimental investigation of their theoretically predicted properties, such as a 2D topological insulating behaviour with a very large bandgap, and the capability to support enhanced thermoelectric performance, topological superconductivity and the near-room-temperature quantum anomalous Hall effect.

Reduced carbon emission estimates from fossil fuel combustion and cement production in China

Abstract: Nearly three-quarters of the growth in global carbon emissions from the burning of fossil fuels and cement production between 2010 and 2012 occurred in China. Yet estimates of Chinese emissions remain subject to large uncertainty; inventories of China's total fossil fuel carbon emissions in 2008 differ by 0.3 gigatonnes of carbon, or 15 per cent. The primary sources of this uncertainty are conflicting estimates of energy consumption and emission factors, the latter being uncertain because of very few actual measurements representative of the mix of Chinese fuels. Here we re-evaluate China's carbon emissions using updated and harmonized energy consumption and clinker production data and two new and comprehensive sets of measured emission factors for Chinese coal. We find that total energy consumption in China was 10 per cent higher in 2000-2012 than the value reported by China's national statistics, that emission factors for Chinese coal are on average 40 per cent lower than the default values recommended by the Intergovernmental Panel on Climate Change, and that emissions from China's cement production are 45 per cent less than recent estimates. Altogether, our revised estimate of China's CO2 emissions from fossil fuel combustion and cement production is 2.49 gigatonnes of carbon (2 standard deviations = ±7.3 per cent) in 2013, which is 14 per cent lower than the emissions reported by other prominent inventories. Over the full period 2000 to 2013, our revised estimates are 2.9 gigatonnes of carbon less than previous estimates of China's cumulative carbon emissions. Our findings suggest that overestimation of China's emissions in 2000-2013 may be larger than China's estimated total forest sink in 1990-2007 (2.66 gigatonnes of carbon) or China's land carbon sink in 2000-2009 (2.6 gigatonnes of carbon).

Transition Metal Dichalcogenides and Beyond: Synthesis, Properties, and Applications of Single- and Few-Layer Nanosheets

Abstract: ConspectusIn the wake of the discovery of the remarkable electronic and physical properties of graphene, a vibrant research area on two-dimensional (2D) layered materials has emerged during the past decade. Transition metal dichalcogenides (TMDs) represent an alternative group of 2D layered materials that differ from the semimetallic character of graphene. They exhibit diverse properties that depend on their composition and can be semiconductors (e.g., MoS2, WS2), semimetals (e.g., WTe2, TiSe2), true metals (e.g., NbS2, VSe2), and superconductors (e.g., NbSe2, TaS2). The properties of TMDs can also be tailored according to the crystalline structure and the number and stacking sequence of layers in their crystals and thin films. For example, 2H-MoS2 is semiconducting, whereas 1T-MoS2 is metallic. Bulk 2H-MoS2 possesses an indirect band gap, but when 2H-MoS2 is exfoliated into monolayers, it exhibits direct electronic and optical band gaps, which leads to enhanced photoluminescence. Therefore, it is importa...

A survey of millimeter wave communications (mmWave) for 5G: opportunities and challenges

Abstract: With the explosive growth of mobile data demand, the fifth generation (5G) mobile network would exploit the enormous amount of spectrum in the millimeter wave (mmWave) bands to greatly increase communication capacity. There are fundamental differences between mmWave communications and existing other communication systems, in terms of high propagation loss, directivity, and sensitivity to blockage. These characteristics of mmWave communications pose several challenges to fully exploit the potential of mmWave communications, including integrated circuits and system design, interference management, spatial reuse, anti-blockage, and dynamics control. To address these challenges, we carry out a survey of existing solutions and standards, and propose design guidelines in architectures and protocols for mmWave communications. We also discuss the potential applications of mmWave communications in the 5G network, including the small cell access, the cellular access, and the wireless backhaul. Finally, we discuss relevant open research issues including the new physical layer technology, software-defined network architecture, measurements of network state information, efficient control mechanisms, and heterogeneous networking, which should be further investigated to facilitate the deployment of mmWave communication systems in the future 5G networks.

PCANet: A Simple Deep Learning Baseline for Image Classification?

Abstract: In this paper, we propose a very simple deep learning network for image classification that is based on very basic data processing components: 1) cascaded principal component analysis (PCA); 2) binary hashing; and 3) blockwise histograms. In the proposed architecture, the PCA is employed to learn multistage filter banks. This is followed by simple binary hashing and block histograms for indexing and pooling. This architecture is thus called the PCA network (PCANet) and can be extremely easily and efficiently designed and learned. For comparison and to provide a better understanding, we also introduce and study two simple variations of PCANet: 1) RandNet and 2) LDANet. They share the same topology as PCANet, but their cascaded filters are either randomly selected or learned from linear discriminant analysis. We have extensively tested these basic networks on many benchmark visual data sets for different tasks, including Labeled Faces in the Wild (LFW) for face verification; the MultiPIE, Extended Yale B, AR, Facial Recognition Technology (FERET) data sets for face recognition; and MNIST for hand-written digit recognition. Surprisingly, for all tasks, such a seemingly naive PCANet model is on par with the state-of-the-art features either prefixed, highly hand-crafted, or carefully learned [by deep neural networks (DNNs)]. Even more surprisingly, the model sets new records for many classification tasks on the Extended Yale B, AR, and FERET data sets and on MNIST variations. Additional experiments on other public data sets also demonstrate the potential of PCANet to serve as a simple but highly competitive baseline for texture classification and object recognition.

Superconductivity above 100 K in single-layer FeSe films on doped SrTiO3

Abstract: Monolayer iron selenide grown on SrTiO3 has recently gained attention due to suggestive evidence it superconducts at high temperature. In situ electrical transport measurements now reveal a transition temperature above 100 K.

Network representation learning with rich text information

Abstract: Representation learning has shown its effectiveness in many tasks such as image classification and text mining. Network representation learning aims at learning distributed vector representation for each vertex in a network, which is also increasingly recognized as an important aspect for network analysis. Most network representation learning methods investigate network structures for learning. In reality, network vertices contain rich information (such as text), which cannot be well applied with algorithmic frameworks of typical representation learning methods. By proving that DeepWalk, a state-of-the-art network representation method, is actually equivalent to matrix factorization (MF), we propose text-associated DeepWalk (TADW). TADW incorporates text features of vertices into network representation learning under the framework of matrix factorization. We evaluate our method and various baseline methods by applying them to the task of multi-class classification of vertices. The experimental results show that, our method outperforms other baselines on all three datasets, especially when networks are noisy and training ratio is small. The source code of this paper can be obtained from https://github.com/albertyang33/TADW.

LHCb detector performance

Abstract: The LHCb detector is a forward spectrometer at the Large Hadron Collider (LHC) at CERN. The experiment is designed for precision measurements of CP violation and rare decays of beauty and charm hadrons. In this paper the performance of the various LHCb sub-detectors and the trigger system are described, using data taken from 2010 to 2012. It is shown that the design criteria of the experiment have been met. The excellent performance of the detector has allowed the LHCb collaboration to publish a wide range of physics results, demonstrating LHCb's unique role, both as a heavy flavour experiment and as a general purpose detector in the forward region.

An efficient molybdenum disulfide/cobalt diselenide hybrid catalyst for electrochemical hydrogen generation.

Abstract: There is substantial research into new catalysts for electroreduction of water. Here, the authors report a robust and active molybdenum disulfide/cobalt diselenide hydrogen evolution catalyst with onset potential of 11 mV and Tafel slope of 36 mV per decade, approaching the activity of platinum.

Observation of $J/\psi p$ Resonances Consistent with Pentaquark States in $\Lambda_b^0 \to J/\psi K^- p$ Decays

Abstract: Observations of exotic structures in the J/ψp channel, that we refer to as pentaquark-charmonium states, in Λ0b→J/ψK−p decays are presented. The data sample corresponds to an integrated luminosity of 3/fb acquired with the LHCb detector from 7 and 8 TeV pp collisions. An amplitude analysis is performed on the three-body final-state that reproduces the two-body mass and angular distributions. To obtain a satisfactory fit of the structures seen in the J/ψp mass spectrum, it is necessary to include two Breit-Wigner amplitudes that each describe a resonant state. The significance of each of these resonances is more than 9 standard deviations. One has a mass of 4380±8±29 MeV and a width of 205±18±86 MeV, while the second is narrower, with a mass of 4449.8±1.7±2.5 MeV and a width of 39±5±19 MeV. The preferred JP assignments are of opposite parity, with one state having spin 3/2 and the other 5/2.

Holey Graphitic Carbon Nitride Nanosheets with Carbon Vacancies for Highly Improved Photocatalytic Hydrogen Production

Abstract: 2D graphitic carbon nitride (GCN) nanosheets have attracted tremendous attention in photocatalysis due to their many intriguing properties. However, the photocatalytic performance of GCN nanosheets is still restricted by the limited active sites and the serious aggregation during the photocatalytic process. Herein, a simple approach to produce holey GCN (HGCN) nanosheets with abundant in-plane holes by thermally treating bulk GCN (BGCN) under an NH3 atmosphere is reported. These formed in-plane holes not only endow GCN nanosheets with more exposed active edges and cross-plane diffusion channels that greatly speed up mass and photogenerated charge transfer, but also provide numerous boundaries and thus decrease the aggregation. Compared to BGCN, the resultant HGCN has a much higher specific surface area of 196 m2 g−1, together with an enlarged bandgap of 2.95 eV. In addition, the HGCN is demonstrated to be self-modified with carbon vacancies that make HGCN show much broader light absorption extending to the near-infrared region, a higher donor density, and remarkably longer lifetime of charge carriers. As such, HGCN has a much higher photocatalytic hydrogen production rate of nearly 20 times the rate of BGCN.

Ultrathin platinum nanowires grown on single-layered nickel hydroxide with high hydrogen evolution activity

Abstract: Design and synthesis of effective electrocatalysts for hydrogen evolution reaction in alkaline environments is critical to reduce energy losses in alkaline water electrolysis. Here we report a hybrid nanomaterial comprising of one-dimensional ultrathin platinum nanowires grown on two-dimensional single-layered nickel hydroxide. Judicious surface chemistry to generate the fully exfoliated nickel hydroxide single layers is explored to be the key for controllable growth of ultrathin platinum nanowires with diameters of about 1.8 nm. Impressively, this hybrid nanomaterial exhibits superior electrocatalytic activity for hydrogen evolution reaction in alkaline solution, which outperforms currently reported catalysts, and the obviously improved catalytic stability. We believe that this work may lead towards the development of single-layered metal hydroxide-based hybrid materials for applications in catalysis and energy conversion.

Weyl semimetal phase in the non-centrosymmetric compound TaAs

Abstract: Experiments show that TaAs is a three-dimensional topological Weyl semimetal. Three-dimensional (3D) topologicalWeyl semimetals (TWSs) represent a state of quantum matter with unusual electronic structures that resemble both a ‘3D graphene’ and a topological insulator. Their electronic structure displays pairs of Weyl points (through which the electronic bands disperse linearly along all three momentum directions) connected by topological surface states, forming a unique arc-like Fermi surface (FS). Each Weyl point is chiral and contains half the degrees of freedom of a Dirac point, and can be viewed as a magnetic monopole in momentum space. By performing angle-resolved photoemission spectroscopy on the non-centrosymmetric compound TaAs, here we report its complete band structure, including the unique Fermi-arc FS and linear bulk band dispersion across the Weyl points, in agreement with the theoretical calculations1,2. This discovery not only confirms TaAs as a 3DTWS, but also provides an ideal platform for realizing exotic physical phenomena (for example, negative magnetoresistance, chiral magnetic effects and the quantum anomalous Hall effect) which may also lead to novel future applications.

Recurrent convolutional neural network for object recognition

Abstract: In recent years, the convolutional neural network (CNN) has achieved great success in many computer vision tasks. Partially inspired by neuroscience, CNN shares many properties with the visual system of the brain. A prominent difference is that CNN is typically a feed-forward architecture while in the visual system recurrent connections are abundant. Inspired by this fact, we propose a recurrent CNN (RCNN) for object recognition by incorporating recurrent connections into each convolutional layer. Though the input is static, the activities of RCNN units evolve over time so that the activity of each unit is modulated by the activities of its neighboring units. This property enhances the ability of the model to integrate the context information, which is important for object recognition. Like other recurrent neural networks, unfolding the RCNN through time can result in an arbitrarily deep network with a fixed number of parameters. Furthermore, the unfolded network has multiple paths, which can facilitate the learning process. The model is tested on four benchmark object recognition datasets: CIFAR-10, CIFAR-100, MNIST and SVHN. With fewer trainable parameters, RCNN outperforms the state-of-the-art models on all of these datasets. Increasing the number of parameters leads to even better performance. These results demonstrate the advantage of the recurrent structure over purely feed-forward structure for object recognition.