Showing papers by "Tsinghua University published in 2017"

Densely Connected Convolutional Networks

Abstract: Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections—one between each layer and its subsequent layer—our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet.

GW170817: observation of gravitational waves from a binary neutron star inspiral

Abstract: On August 17, 2017 at 12∶41:04 UTC the Advanced LIGO and Advanced Virgo gravitational-wave detectors made their first observation of a binary neutron star inspiral. The signal, GW170817, was detected with a combined signal-to-noise ratio of 32.4 and a false-alarm-rate estimate of less than one per 8.0×10^{4} years. We infer the component masses of the binary to be between 0.86 and 2.26 M_{⊙}, in agreement with masses of known neutron stars. Restricting the component spins to the range inferred in binary neutron stars, we find the component masses to be in the range 1.17-1.60 M_{⊙}, with the total mass of the system 2.74_{-0.01}^{+0.04}M_{⊙}. The source was localized within a sky region of 28 deg^{2} (90% probability) and had a luminosity distance of 40_{-14}^{+8} Mpc, the closest and most precisely localized gravitational-wave signal yet. The association with the γ-ray burst GRB 170817A, detected by Fermi-GBM 1.7 s after the coalescence, corroborates the hypothesis of a neutron star merger and provides the first direct evidence of a link between these mergers and short γ-ray bursts. Subsequent identification of transient counterparts across the electromagnetic spectrum in the same location further supports the interpretation of this event as a neutron star merger. This unprecedented joint gravitational and electromagnetic observation provides insight into astrophysics, dense matter, gravitation, and cosmology.

Toward Safe Lithium Metal Anode in Rechargeable Batteries: A Review.

Abstract: The lithium metal battery is strongly considered to be one of the most promising candidates for high-energy-density energy storage devices in our modern and technology-based society. However, uncontrollable lithium dendrite growth induces poor cycling efficiency and severe safety concerns, dragging lithium metal batteries out of practical applications. This review presents a comprehensive overview of the lithium metal anode and its dendritic lithium growth. First, the working principles and technical challenges of a lithium metal anode are underscored. Specific attention is paid to the mechanistic understandings and quantitative models for solid electrolyte interphase (SEI) formation, lithium dendrite nucleation, and growth. On the basis of previous theoretical understanding and analysis, recently proposed strategies to suppress dendrite growth of lithium metal anode and some other metal anodes are reviewed. A section dedicated to the potential of full-cell lithium metal batteries for practical applicatio...

Deformable Convolutional Networks

Abstract: Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules. In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting the spatial sampling locations in the modules with additional offsets and learning the offsets from the target tasks, without additional supervision. The new modules can readily replace their plain counterparts in existing CNNs and can be easily trained end-to-end by standard back-propagation, giving rise to deformable convolutional networks. Extensive experiments validate the performance of our approach. For the first time, we show that learning dense spatial transformation in deep CNNs is effective for sophisticated vision tasks such as object detection and semantic segmentation. The code is released at https://github.com/msracver/Deformable-ConvNets.

Gravitational Waves and Gamma-Rays from a Binary Neutron Star Merger: GW170817 and GRB 170817A

Abstract: On 2017 August 17, the gravitational-wave event GW170817 was observed by the Advanced LIGO and Virgo detectors, and the gamma-ray burst (GRB) GRB 170817A was observed independently by the Fermi Gamma-ray Burst Monitor, and the Anti-Coincidence Shield for the Spectrometer for the International Gamma-Ray Astrophysics Laboratory. The probability of the near-simultaneous temporal and spatial observation of GRB 170817A and GW170817 occurring by chance is $5.0\times {10}^{-8}$. We therefore confirm binary neutron star mergers as a progenitor of short GRBs. The association of GW170817 and GRB 170817A provides new insight into fundamental physics and the origin of short GRBs. We use the observed time delay of $(+1.74\pm 0.05)\,{\rm{s}}$ between GRB 170817A and GW170817 to: (i) constrain the difference between the speed of gravity and the speed of light to be between $-3\times {10}^{-15}$ and $+7\times {10}^{-16}$ times the speed of light, (ii) place new bounds on the violation of Lorentz invariance, (iii) present a new test of the equivalence principle by constraining the Shapiro delay between gravitational and electromagnetic radiation. We also use the time delay to constrain the size and bulk Lorentz factor of the region emitting the gamma-rays. GRB 170817A is the closest short GRB with a known distance, but is between 2 and 6 orders of magnitude less energetic than other bursts with measured redshift. A new generation of gamma-ray detectors, and subthreshold searches in existing detectors, will be essential to detect similar short bursts at greater distances. Finally, we predict a joint detection rate for the Fermi Gamma-ray Burst Monitor and the Advanced LIGO and Virgo detectors of 0.1–1.4 per year during the 2018–2019 observing run and 0.3–1.7 per year at design sensitivity.

Residual Attention Network for Image Classification

Abstract: In this work, we propose Residual Attention Network, a convolutional neural network using attention mechanism which can incorporate with state-of-art feed forward network architecture in an end-to-end training fashion. Our Residual Attention Network is built by stacking Attention Modules which generate attention-aware features. The attention-aware features from different modules change adaptively as layers going deeper. Inside each Attention Module, bottom-up top-down feedforward structure is used to unfold the feedforward and feedback attention process into a single feedforward process. Importantly, we propose attention residual learning to train very deep Residual Attention Networks which can be easily scaled up to hundreds of layers. Extensive analyses are conducted on CIFAR-10 and CIFAR-100 datasets to verify the effectiveness of every module mentioned above. Our Residual Attention Network achieves state-of-the-art object recognition performance on three benchmark datasets including CIFAR-10 (3.90% error), CIFAR-100 (20.45% error) and ImageNet (4.8% single model and single crop, top-5 error). Note that, our method achieves 0.6% top-1 accuracy improvement with 46% trunk depth and 69% forward FLOPs comparing to ResNet-200. The experiment also demonstrates that our network is robust against noisy labels.

Multi-view 3D Object Detection Network for Autonomous Driving

Abstract: This paper aims at high-accuracy 3D object detection in autonomous driving scenario. We propose Multi-View 3D networks (MV3D), a sensory-fusion framework that takes both LIDAR point cloud and RGB images as input and predicts oriented 3D bounding boxes. We encode the sparse 3D point cloud with a compact multi-view representation. The network is composed of two subnetworks: one for 3D object proposal generation and another for multi-view feature fusion. The proposal network generates 3D candidate boxes efficiently from the birds eye view representation of 3D point cloud. We design a deep fusion scheme to combine region-wise features from multiple views and enable interactions between intermediate layers of different paths. Experiments on the challenging KITTI benchmark show that our approach outperforms the state-of-the-art by around 25% and 30% AP on the tasks of 3D localization and 3D detection. In addition, for 2D detection, our approach obtains 14.9% higher AP than the state-of-the-art on the hard data among the LIDAR-based methods.

GW170104: Observation of a 50-Solar-Mass Binary Black Hole Coalescence at Redshift 0.2

Abstract: We describe the observation of GW170104, a gravitational-wave signal produced by the coalescence of a pair of stellar-mass black holes. The signal was measured on January 4, 2017 at 10∶11:58.6 UTC by the twin advanced detectors of the Laser Interferometer Gravitational-Wave Observatory during their second observing run, with a network signal-to-noise ratio of 13 and a false alarm rate less than 1 in 70 000 years. The inferred component black hole masses are 31.2^(8.4) _(−6.0)M_⊙ and 19.4^(5.3)_( −5.9)M_⊙ (at the 90% credible level). The black hole spins are best constrained through measurement of the effective inspiral spin parameter, a mass-weighted combination of the spin components perpendicular to the orbital plane, χ_(eff) = −0.12^(0.21)_( −0.30). This result implies that spin configurations with both component spins positively aligned with the orbital angular momentum are disfavored. The source luminosity distance is 880^(450)_(−390) Mpc corresponding to a redshift of z = 0.18^(0.08)_( −0.07) . We constrain the magnitude of modifications to the gravitational-wave dispersion relation and perform null tests of general relativity. Assuming that gravitons are dispersed in vacuum like massive particles, we bound the graviton mass to m_g ≤ 7.7 × 10^(−23) eV/c^2. In all cases, we find that GW170104 is consistent with general relativity.

GW170814: A three-detector observation of gravitational waves from a binary black hole coalescence

Abstract: On August 14, 2017 at 10∶30:43 UTC, the Advanced Virgo detector and the two Advanced LIGO detectors coherently observed a transient gravitational-wave signal produced by the coalescence of two stellar mass black holes, with a false-alarm rate of ≲1 in 27 000 years. The signal was observed with a three-detector network matched-filter signal-to-noise ratio of 18. The inferred masses of the initial black holes are 30.5-3.0+5.7M⊙ and 25.3-4.2+2.8M⊙ (at the 90% credible level). The luminosity distance of the source is 540-210+130 Mpc, corresponding to a redshift of z=0.11-0.04+0.03. A network of three detectors improves the sky localization of the source, reducing the area of the 90% credible region from 1160 deg2 using only the two LIGO detectors to 60 deg2 using all three detectors. For the first time, we can test the nature of gravitational-wave polarizations from the antenna response of the LIGO-Virgo network, thus enabling a new class of phenomenological tests of gravity.

Learning Efficient Convolutional Networks through Network Slimming

Abstract: The deployment of deep convolutional neural networks (CNNs) in many real world applications is largely hindered by their high computational cost. In this paper, we propose a novel learning scheme for CNNs to simultaneously 1) reduce the model size; 2) decrease the run-time memory footprint; and 3) lower the number of computing operations, without compromising accuracy. This is achieved by enforcing channel-level sparsity in the network in a simple but effective way. Different from many existing approaches, the proposed method directly applies to modern CNN architectures, introduces minimum overhead to the training process, and requires no special software/hardware accelerators for the resulting models. We call our approach network slimming, which takes wide and large networks as input models, but during training insignificant channels are automatically identified and pruned afterwards, yielding thin and compact models with comparable accuracy. We empirically demonstrate the effectiveness of our approach with several state-of-the-art CNN models, including VGGNet, ResNet and DenseNet, on various image classification datasets. For VGGNet, a multi-pass version of network slimming gives a 20× reduction in model size and a 5× reduction in computing operations.

Deformable Convolutional Networks

Abstract: Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in its building modules. In this work, we introduce two new modules to enhance the transformation modeling capacity of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting the spatial sampling locations in the modules with additional offsets and learning the offsets from target tasks, without additional supervision. The new modules can readily replace their plain counterparts in existing CNNs and can be easily trained end-to-end by standard back-propagation, giving rise to deformable convolutional networks. Extensive experiments validate the effectiveness of our approach on sophisticated vision tasks of object detection and semantic segmentation. The code would be released.

Deep transfer learning with joint adaptation networks

Abstract: Deep networks have been successfully applied to learn transferable features for adapting models from a source domain to a different target domain. In this paper, we present joint adaptation networks (JAN), which learn a transfer network by aligning the joint distributions of multiple domain-specific layers across domains based on a joint maximum mean discrepancy (JMMD) criterion. Adversarial training strategy is adopted to maximize JMMD such that the distributions of the source and target domains are made more distinguishable. Learning can be performed by stochastic gradient descent with the gradients computed by back-propagation in linear-time. Experiments testify that our model yields state of the art results on standard datasets.

Isolated Single Iron Atoms Anchored on N-Doped Porous Carbon as an Efficient Electrocatalyst for the Oxygen Reduction Reaction

Abstract: The development of low-cost, efficient, and stable electrocatalysts for the oxygen reduction reaction (ORR) is desirable but remains a great challenge. Herein, we made a highly reactive and stable isolated single-atom Fe/N-doped porous carbon (ISA Fe/CN) catalyst with Fe loading up to 2.16 wt %. The catalyst showed excellent ORR performance with a half-wave potential (E1/2) of 0.900 V, which outperformed commercial Pt/C and most non-precious-metal catalysts reported to date. Besides exceptionally high kinetic current density (Jk) of 37.83 mV cm−2 at 0.85 V, it also had a good methanol tolerance and outstanding stability. Experiments demonstrated that maintaining the Fe as isolated atoms and incorporating nitrogen was essential to deliver the high performance. First principle calculations further attributed the high reactivity to the high efficiency of the single Fe atoms in transporting electrons to the adsorbed OH species.

A Point Set Generation Network for 3D Object Reconstruction from a Single Image

Abstract: Generation of 3D data by deep neural network has been attracting increasing attention in the research community. The majority of extant works resort to regular representations such as volumetric grids or collection of images, however, these representations obscure the natural invariance of 3D shapes under geometric transformations, and also suffer from a number of other issues. In this paper we address the problem of 3D reconstruction from a single image, generating a straight-forward form of output – point cloud coordinates. Along with this problem arises a unique and interesting issue, that the groundtruth shape for an input image may be ambiguous. Driven by this unorthordox output form and the inherent ambiguity in groundtruth, we design architecture, loss function and learning paradigm that are novel and effective. Our final solution is a conditional shape sampler, capable of predicting multiple plausible 3D point clouds from an input image. In experiments not only can our system outperform state-of-the-art methods on single image based 3D reconstruction benchmarks, but it also shows strong performance for 3D shape completion and promising ability in making multiple plausible predictions.

Review on High-Loading and High-Energy Lithium–Sulfur Batteries

Abstract: Owing to high specific energy, low cost, and environmental friendliness, lithium–sulfur (Li–S) batteries hold great promise to meet the increasing demand for advanced energy storage beyond portable electronics, and to mitigate environmental problems. However, the application of Li–S batteries is challenged by several obstacles, including their short life and low sulfur utilization, which become more serious when sulfur loading is increased to the practically accepted level above 3–5 mg cm−2. More and more efforts have been made recently to overcome the barriers toward commercially viable Li–S batteries with a high sulfur loading. This review highlights the recent progress in high-sulfur-loading Li–S batteries enabled by hierarchical design principles at multiscale. Particularly, basic insights into the interfacial reactions, strategies for mesoscale assembly, unique architectures, and configurational innovation in the cathode, anode, and separator are under specific concerns. Hierarchy in the multiscale design is proposed to guide the future development of high-sulfur-loading Li–S batteries.

Residual Attention Network for Image Classification

Abstract: In this work, we propose "Residual Attention Network", a convolutional neural network using attention mechanism which can incorporate with state-of-art feed forward network architecture in an end-to-end training fashion. Our Residual Attention Network is built by stacking Attention Modules which generate attention-aware features. The attention-aware features from different modules change adaptively as layers going deeper. Inside each Attention Module, bottom-up top-down feedforward structure is used to unfold the feedforward and feedback attention process into a single feedforward process. Importantly, we propose attention residual learning to train very deep Residual Attention Networks which can be easily scaled up to hundreds of layers. Extensive analyses are conducted on CIFAR-10 and CIFAR-100 datasets to verify the effectiveness of every module mentioned above. Our Residual Attention Network achieves state-of-the-art object recognition performance on three benchmark datasets including CIFAR-10 (3.90% error), CIFAR-100 (20.45% error) and ImageNet (4.8% single model and single crop, top-5 error). Note that, our method achieves 0.6% top-1 accuracy improvement with 46% trunk depth and 69% forward FLOPs comparing to ResNet-200. The experiment also demonstrates that our network is robust against noisy labels.

More Reliable Lithium-Sulfur Batteries: Status, Solutions and Prospects.

Abstract: Lithium-sulfur (Li-S) batteries have attracted tremendous interest because of their high theoretical energy density and cost effectiveness. The target of Li-S battery research is to produce batteries with a high useful energy density that at least outperforms state-of-the-art lithium-ion batteries. However, due to an intrinsic gap between fundamental research and practical applications, the outstanding electrochemical results obtained in most Li-S battery studies indeed correspond to low useful energy densities and are not really suitable for practical requirements. The Li-S battery is a complex device and its useful energy density is determined by a number of design parameters, most of which are often ignored, leading to the failure to meet commercial requirements. The purpose of this review is to discuss how to pave the way for reliable Li-S batteries. First, the current research status of Li-S batteries is briefly reviewed based on statistical information obtained from literature. This includes an analysis of how the various parameters influence the useful energy density and a summary of existing problems in the current Li-S battery research. Possible solutions and some concerns regarding the construction of reliable Li-S batteries are comprehensively discussed. Finally, insights are offered on the future directions and prospects in Li-S battery field.

GW170608: Observation of a 19 solar-mass binary black hole coalescence

Abstract: On 2017 June 8 at 02:01:16.49 UTC, a gravitational-wave (GW) signal from the merger of two stellar-mass black holes was observed by the two Advanced Laser Interferometer Gravitational-Wave Observatory detectors with a network signal-to-noise ratio of 13. This system is the lightest black hole binary so far observed, with component masses of ${12}_{-2}^{+7}\,{M}_{\odot }$ and ${7}_{-2}^{+2}\,{M}_{\odot }$ (90% credible intervals). These lie in the range of measured black hole masses in low-mass X-ray binaries, thus allowing us to compare black holes detected through GWs with electromagnetic observations. The source's luminosity distance is ${340}_{-140}^{+140}\,\mathrm{Mpc}$, corresponding to redshift ${0.07}_{-0.03}^{+0.03}$. We verify that the signal waveform is consistent with the predictions of general relativity.

Nanostructured Metal Oxides and Sulfides for Lithium-Sulfur Batteries

Abstract: Lithium-sulfur (Li-S) batteries with high energy density and long cycle life are considered to be one of the most promising next-generation energy-storage systems beyond routine lithium-ion batteries. Various approaches have been proposed to break down technical barriers in Li-S battery systems. The use of nanostructured metal oxides and sulfides for high sulfur utilization and long life span of Li-S batteries is reviewed here. The relationships between the intrinsic properties of metal oxide/sulfide hosts and electrochemical performances of Li-S batteries are discussed. Nanostructured metal oxides/sulfides hosts used in solid sulfur cathodes, separators/interlayers, lithium-metal-anode protection, and lithium polysulfides batteries are discussed respectively. Prospects for the future developments of Li-S batteries with nanostructured metal oxides/sulfides are also discussed.

Sloan Digital Sky Survey IV: Mapping the Milky Way, Nearby Galaxies and the Distant Universe

Abstract: We describe the Sloan Digital Sky Survey IV (SDSS-IV), a project encompassing three major spectroscopic programs. The Apache Point Observatory Galactic Evolution Experiment 2 (APOGEE-2) is observing hundreds of thousands of Milky Way stars at high resolution and high signal-to-noise ratios in the near-infrared. The Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey is obtaining spatially resolved spectroscopy for thousands of nearby galaxies (median $z\sim 0.03$). The extended Baryon Oscillation Spectroscopic Survey (eBOSS) is mapping the galaxy, quasar, and neutral gas distributions between $z\sim 0.6$ and 3.5 to constrain cosmology using baryon acoustic oscillations, redshift space distortions, and the shape of the power spectrum. Within eBOSS, we are conducting two major subprograms: the SPectroscopic IDentification of eROSITA Sources (SPIDERS), investigating X-ray AGNs and galaxies in X-ray clusters, and the Time Domain Spectroscopic Survey (TDSS), obtaining spectra of variable sources. All programs use the 2.5 m Sloan Foundation Telescope at the Apache Point Observatory; observations there began in Summer 2014. APOGEE-2 also operates a second near-infrared spectrograph at the 2.5 m du Pont Telescope at Las Campanas Observatory, with observations beginning in early 2017. Observations at both facilities are scheduled to continue through 2020. In keeping with previous SDSS policy, SDSS-IV provides regularly scheduled public data releases; the first one, Data Release 13, was made available in 2016 July.

Fluoroethylene Carbonate Additives to Render Uniform Li Deposits in Lithium Metal Batteries

Abstract: Lithium (Li) metal has been considered as an important substitute for the graphite anode to further boost the energy density of Li-ion batteries. However, Li dendrite growth during Li plating/stripping causes safety concern and poor lifespan of Li metal batteries (LMB). Herein, fluoroethylene carbonate (FEC) additives are used to form a LiF-rich solid electrolyte interphase (SEI). The FEC-induced SEI layer is compact and stable, and thus beneficial to obtain a uniform morphology of Li deposits. This uniform and dendrite-free morphology renders a significantly improved Coulombic efficiency of 98% within 100 cycles in a Li | Cu half-cell. When the FEC-protected Li metal anode matches a high-loading LiNi0.5Co0.2Mn0.3O2 (NMC) cathode (12 mg cm−2), a high initial capacity of 154 mAh g−1 (1.9 mAh cm−2) at 180.0 mA g−1 is obtained. This LMB with conversion-type Li metal anode and intercalation-type NMC cathode affords an emerging energy storage system to probe the energy chemistry of Li metal protection and demonstrates the material engineering of batteries with very high energy density.

Stochastic Community Assembly: Does It Matter in Microbial Ecology?

Abstract: Understanding the mechanisms controlling community diversity, functions, succession, and biogeography is a central, but poorly understood, topic in ecology, particularly in microbial ecology. Although stochastic processes are believed to play nonnegligible roles in shaping community structure, their importance relative to deterministic processes is hotly debated. The importance of ecological stochasticity in shaping microbial community structure is far less appreciated. Some of the main reasons for such heavy debates are the difficulty in defining stochasticity and the diverse methods used for delineating stochasticity. Here, we provide a critical review and synthesis of data from the most recent studies on stochastic community assembly in microbial ecology. We then describe both stochastic and deterministic components embedded in various ecological processes, including selection, dispersal, diversification, and drift. We also describe different approaches for inferring stochasticity from observational diversity patterns and highlight experimental approaches for delineating ecological stochasticity in microbial communities. In addition, we highlight research challenges, gaps, and future directions for microbial community assembly research.

Design of N-Coordinated Dual-Metal Sites: A Stable and Active Pt-Free Catalyst for Acidic Oxygen Reduction Reaction

Abstract: We develop a host-guest strategy to construct an electrocatalyst with Fe-Co dual sites embedded on N-doped porous carbon and demonstrate its activity for oxygen reduction reaction in acidic electrolyte. Our catalyst exhibits superior oxygen reduction reaction performance, with comparable onset potential (Eonset, 1.06 vs 1.03 V) and half-wave potential (E1/2, 0.863 vs 0.858 V) than commercial Pt/C. The fuel cell test reveals (Fe,Co)/N-C outperforms most reported Pt-free catalysts in H2/O2 and H2/air. In addition, this cathode catalyst with dual metal sites is stable in a long-term operation with 50 000 cycles for electrode measurement and 100 h for H2/air single cell operation. Density functional theory calculations reveal the dual sites is favored for activation of O-O, crucial for four-electron oxygen reduction.

Large Kernel Matters — Improve Semantic Segmentation by Global Convolutional Network

Abstract: One of recent trends [31, 32, 14] in network architecture design is stacking small filters (e.g., 1x1 or 3x3) in the entire network because the stacked small filters is more efficient than a large kernel, given the same computational complexity. However, in the field of semantic segmentation, where we need to perform dense per-pixel prediction, we find that the large kernel (and effective receptive field) plays an important role when we have to perform the classification and localization tasks simultaneously. Following our design principle, we propose a Global Convolutional Network to address both the classification and localization issues for the semantic segmentation. We also suggest a residual-based boundary refinement to further refine the object boundaries. Our approach achieves state-of-art performance on two public benchmarks and significantly outperforms previous results, 82.2% (vs 80.2%) on PASCAL VOC 2012 dataset and 76.9% (vs 71.8%) on Cityscapes dataset.

Historical (1750–2014) anthropogenic emissions of reactive gases and aerosols from the Community Emissions Data System (CEDS)

Abstract: . We present a new data set of annual historical (1750–2014) anthropogenic chemically reactive gases (CO, CH4, NH3, NOx, SO2, NMVOCs), carbonaceous aerosols (black carbon – BC, and organic carbon – OC), and CO2 developed with the Community Emissions Data System (CEDS). We improve upon existing inventories with a more consistent and reproducible methodology applied to all emission species, updated emission factors, and recent estimates through 2014. The data system relies on existing energy consumption data sets and regional and country-specific inventories to produce trends over recent decades. All emission species are consistently estimated using the same activity data over all time periods. Emissions are provided on an annual basis at the level of country and sector and gridded with monthly seasonality. These estimates are comparable to, but generally slightly higher than, existing global inventories. Emissions over the most recent years are more uncertain, particularly in low- and middle-income regions where country-specific emission inventories are less available. Future work will involve refining and updating these emission estimates, estimating emissions' uncertainty, and publication of the system as open-source software.

Ionic Exchange of Metal–Organic Frameworks to Access Single Nickel Sites for Efficient Electroreduction of CO2

Abstract: Single-atom catalysts often exhibit unexpected catalytic activity for many important chemical reactions because of their unique electronic and geometric structures with respect to their bulk counterparts. Herein we adopt metal–organic frameworks (MOFs) to assist the preparation of a catalyst containing single Ni sites for efficient electroreduction of CO2. The synthesis is based on ionic exchange between Zn nodes and adsorbed Ni ions within the cavities of the MOF. This single-atom catalyst exhibited an excellent turnover frequency for electroreduction of CO2 (5273 h–1), with a Faradaic efficiency for CO production of over 71.9% and a current density of 10.48 mA cm–2 at an overpotential of 0.89 V. Our findings present some guidelines for the rational design and accurate modulation of nanostructured catalysts at the atomic scale.

Lithiophilic Sites in Doped Graphene Guide Uniform Lithium Nucleation for Dendrite‐Free Lithium Metal Anodes

Abstract: Lithium (Li) metal is the most promising electrode for next-generation rechargeable batteries. However, the challenges induced by Li dendrites on a working Li metal anode hinder the practical applications of Li metal batteries. Herein, nitrogen (N) doped graphene was adopted as the Li plating matrix to regulate Li metal nucleation and suppress dendrite growth. The N-containing functional groups, such as pyridinic and pyrrolic nitrogen in the N-doped graphene, are lithiophilic, which guide the metallic Li nucleation causing the metal to distribute uniformly on the anode surface. As a result, the N-doped graphene modified Li metal anode exhibits a dendrite-free morphology during repeated Li plating and demonstrates a high Coulombic efficiency of 98 % for near 200 cycles.

MIX: a mosaic Asian anthropogenic emission inventory under the international collaboration framework of the MICS-Asia and HTAP

Abstract: . The MIX inventory is developed for the years 2008 and 2010 to support the Model Inter-Comparison Study for Asia (MICS-Asia) and the Task Force on Hemispheric Transport of Air Pollution (TF HTAP) by a mosaic of up-to-date regional emission inventories. Emissions are estimated for all major anthropogenic sources in 29 countries and regions in Asia. We conducted detailed comparisons of different regional emission inventories and incorporated the best available ones for each region into the mosaic inventory at a uniform spatial and temporal resolution. Emissions are aggregated to five anthropogenic sectors: power, industry, residential, transportation, and agriculture. We estimate the total Asian emissions of 10 species in 2010 as follows: 51.3 Tg SO2, 52.1 Tg NOx, 336.6 Tg CO, 67.0 Tg NMVOC (non-methane volatile organic compounds), 28.8 Tg NH3, 31.7 Tg PM10, 22.7 Tg PM2.5, 3.5 Tg BC, 8.3 Tg OC, and 17.3 Pg CO2. Emissions from China and India dominate the emissions of Asia for most of the species. We also estimated Asian emissions in 2006 using the same methodology of MIX. The relative change rates of Asian emissions for the period of 2006–2010 are estimated as follows: −8.1 % for SO2, +19.2 % for NOx, +3.9 % for CO, +15.5 % for NMVOC, +1.7 % for NH3, −3.4 % for PM10, −1.6 % for PM2.5, +5.5 % for BC, +1.8 % for OC, and +19.9 % for CO2. Model-ready speciated NMVOC emissions for SAPRC-99 and CB05 mechanisms were developed following a profile-assignment approach. Monthly gridded emissions at a spatial resolution of 0.25° × 0.25° are developed and can be accessed from http://www.meicmodel.org/dataset-mix .

Conditional Adversarial Domain Adaptation

Abstract: Adversarial learning has been embedded into deep networks to learn disentangled and transferable representations for domain adaptation. Existing adversarial domain adaptation methods may not effectively align different domains of multimodal distributions native in classification problems. In this paper, we present conditional adversarial domain adaptation, a principled framework that conditions the adversarial adaptation models on discriminative information conveyed in the classifier predictions. Conditional domain adversarial networks (CDANs) are designed with two novel conditioning strategies: multilinear conditioning that captures the cross-covariance between feature representations and classifier predictions to improve the discriminability, and entropy conditioning that controls the uncertainty of classifier predictions to guarantee the transferability. With theoretical guarantees and a few lines of codes, the approach has exceeded state-of-the-art results on five datasets.