Showing papers by "Qian Zhang published in 2020"
Helmholtz Centre for Environmental Research - UFZ1, Catalan Institute for Water Research2, University of Girona3, University of Barcelona4, Leibniz Association5, University of Potsdam6, Griffith University7, Abant Izzet Baysal University8, Universidade Federal de Juiz de Fora9, Charles Sturt University10, Aarhus University11, Radboud University Nijmegen12, Uppsala University13, University of Murcia14, University of Innsbruck15, Federal University of Rio Grande do Norte16, Estonian University of Life Sciences17, Ewha Womans University18, Chinese Academy of Sciences19, National Scientific and Technical Research Council20, University of the Basque Country21, UNESCO-IHE Institute for Water Education22, IRSA23, Umeå University24
TL;DR: A global survey covering 196 dry inland waters shows that their CO2 emissions share fundamental drivers and constitute a substantial fraction of the carbon cycled by inland waters, increasing current inland water carbon flux estimates by 6%.
Abstract: Many inland waters exhibit complete or partial desiccation, or have vanished due to global change, exposing sediments to the atmosphere. Yet, data on carbon dioxide (CO2) emissions from these sedim ...
474 citations
07 Feb 2020
TL;DR: An end-to-end speech recognition model with Transformer encoders that can be used in a streaming speech recognition system and shows that the full attention version of the model beats the-state-of-the art accuracy on the LibriSpeech benchmarks.
Abstract: In this paper we present an end-to-end speech recognition model with Transformer encoders that can be used in a streaming speech recognition system. Transformer computation blocks based on self-attention are used to encode both audio and label sequences independently. The activations from both audio and label encoders are combined with a feed-forward layer to compute a probability distribution over the label space for every combination of acoustic frame position and label history. This is similar to the Recurrent Neural Network Transducer (RNN-T) model, which uses RNNs for information encoding instead of Transformer encoders. The model is trained with the RNN-T loss well-suited to streaming decoding. We present results on the LibriSpeech dataset showing that limiting the left context for self-attention in the Transformer layers makes decoding computationally tractable for streaming, with only a slight degradation in accuracy. We also show that the full attention version of our model beats the-state-of-the art accuracy on the LibriSpeech benchmarks. Our results also show that we can bridge the gap between full attention and limited attention versions of our model by attending to a limited number of future frames.
382 citations
72 citations
TL;DR: In this paper, a ruthenium sulfonate ligand is synthesized to coordinate black phosphorus (BP) nanosheets, and the resulting RuL3@BP is added into epoxy resin (EP) to fabricate EP nanocomposites.
63 citations
TL;DR: It is confirmed that strong phase fluctuation is an important character in the 2D iron-basedsuperconductors as widely observed in high-T_{c} cuprate superconductors.
Abstract: Superconductivity arises from two distinct quantum phenomena: electron pairing and long-range phase coherence. In conventional superconductors, the two quantum phenomena generally take place simultaneously, while in the underdoped high- ${T}_{c}$ cuprate superconductors, the electron pairing occurs at higher temperature than the long-range phase coherence. Recently, whether electron pairing is also prior to long-range phase coherence in single-layer FeSe film on ${\mathrm{SrTiO}}_{3}$ substrate is under debate. Here, by measuring Knight shift and nuclear spin-lattice relaxation rate, we unambiguously reveal a pseudogap behavior below ${T}_{p}\ensuremath{\sim}60\text{ }\text{ }\mathrm{K}$ in two kinds of layered FeSe-based superconductors with quasi2D nature. In the pseudogap regime, a weak diamagnetic signal and a remarkable Nernst effect are also observed, which indicates that the observed pseudogap behavior is related to superconducting fluctuations. These works confirm that strong phase fluctuation is an important character in the 2D iron-based superconductors as widely observed in high-${T}_{c}$ cuprate superconductors.
48 citations
TL;DR: In this paper, a high conversion efficiency of up to 9 % was achieved with a reliable Mg3.2Sb1.5Bi0.49Te0.01/304 stainless steel junction at a temperature difference of 370 K (Th = 673 K, Tc = 303 K).
47 citations
TL;DR: In this paper, bioinspired dopamine chemistry was used to form thin polydopamine (PDA) coatings on the surface of nanodiamond (ND) particles (ND@PDA).
38 citations
Posted Content•
TL;DR: This paper presents a Transformer-Transducer model architecture and a training technique to unify streaming and non-streaming speech recognition models into one model, and shows that with limited right context and small additional latency at the end of decoding, this model can achieve similar accuracy with models using unlimited audio right context.
Abstract: In this paper we present a Transformer-Transducer model architecture and a training technique to unify streaming and non-streaming speech recognition models into one model. The model is composed of a stack of transformer layers for audio encoding with no lookahead or right context and an additional stack of transformer layers on top trained with variable right context. In inference time, the context length for the variable context layers can be changed to trade off the latency and the accuracy of the model. We also show that we can run this model in a Y-model architecture with the top layers running in parallel in low latency and high latency modes. This allows us to have streaming speech recognition results with limited latency and delayed speech recognition results with large improvements in accuracy (20% relative improvement for voice-search task). We show that with limited right context (1-2 seconds of audio) and small additional latency (50-100 milliseconds) at the end of decoding, we can achieve similar accuracy with models using unlimited audio right context. We also present optimizations for audio and label encoders to speed up the inference in streaming and non-streaming speech decoding.
35 citations
34 citations
TL;DR: It is reported that NST activity is modulated by the NAC domain transcription factor XYLEM NAC DOMAIN1 (XND1) during plant growth and XND1 is identified as an NST-interacting protein that modulates NST1 activity.
Abstract: In Arabidopsis, secondary cell walls (SCW) are formed in fiber cells and vessel cells in vascular tissue for providing plants with mechanical strength and channels for the long distance transportation of water and nutrients. NAC SECONDARY WALL THICKENING PROMOTING FACTOR1 (NST1) acts as a key gene for the initiation of SCW formation through a hierarchical transcription network. In this study, we report that NST activity is modulated by the NAC domain transcription factor XYLEM NAC DOMAIN1 (XND1) during plant growth. Using yeast two-hybrid screening and in vivo protein interaction analysis, XND1 was identified as an NST-interacting protein that modulates NST1 activity. XND1 and NST1 were co-localized in the nucleus and the interaction of XND1 with NST1 resulted in inhibition of NST1 transactivation activity. In the process of inflorescence growth, XND1 was expressed with a similar pattern to NST1. Up-regulation of XND1 in fiber cells repressed SCW formation. The study demonstrates that NST1 activity is modulated by XND1 in the regulation of secondary cell walls formation.
33 citations
TL;DR: In this article, n-type Sn1-xPrxSe polycrystals are prepared by ball milling and hot pressing, and a maximum ZT value of 0.7 at 773 K is achieved in Sn0.97Pr0.03Se due to the enhanced electron concentration and electrical conductivity resulting from Pr doping at the Sn site.
TL;DR: Recently, the rational design and fabrication of Prussian blue (PB) cathodes with a unique morphology has been regarded as a promising way to improve the electrochemical properties for sodium ion b... as discussed by the authors.
Abstract: Recently, the rational design and fabrication of Prussian blue (PB) cathodes with a unique morphology has been regarded as a promising way to improve the electrochemical properties for sodium ion b...
TL;DR: Differences in discharge characteristics including visual images, voltage-current waveforms, Lissajous figures, and temperature variation, were determined between the DDBD and SDBD reactors to evaluate the toluene abatement performance of the two reactors.
TL;DR: In this paper, a facile strategy to produce lignin microspheres (LMS) with loading of benzotriazole (BTA) inhibitors, which were released as the pH changed, was proposed.
Abstract: In this work, we offered a facile strategy to produce lignin microspheres (LMS) with loading of benzotriazole (BTA) inhibitors, which were released as the pH changed. The morphological and composit...
25 Oct 2020
TL;DR: This paper investigates a streamable end-to-end multilingual system based on the Transformer Transducer and proposes several techniques for adapting the self-attention architecturebased on the language id, and analyses the trade-offs of each method with regards to quality gains and number of additional parameters introduced.
Abstract: Multilingual automatic speech recognition systems can transcribe utterances from different languages. These systems are attractive from different perspectives: they can provide quality improvements, specially for lower resource languages, and simplify the training and deployment procedure. End-to-end speech recognition has further simplified multilingual modeling as one model, instead of several components of a classical system, have to be unified. In this paper, we investigate a streamable end-to-end multilingual system based on the Transformer Transducer [1]. We propose several techniques for adapting the self-attention architecture based on the language id. We analyze the trade-offs of each method with regards to quality gains and number of additional parameters introduced. We conduct experiments in a real-world task consisting of five languages. Our experimental results demonstrate ∼8% to ∼20% relative gain over the baseline multilingual model.
TL;DR: In this article, a series of ZrBeSi-type Zintl-phase compressive thermoelectric materials have been used to construct a ZnL-glass-like material.
Abstract: Zintl-phase thermoelectric materials have garnered a lot of attention because of their intrinsic electron-crystal and phonon-glass structures. In this work, a series of ZrBeSi-type Zintl-phase comp...
TL;DR: In this paper, the effects of polysiloxane-based polyurethane/lignin elastomers on thermal properties, mechanical properties, surface hydrophobicity, and cytotoxicity in vitro are investigated.
Abstract: Polysiloxane-based polyurethane/lignin elastomers are synthesized via pre-polymerization. Isophorone diisocyanate (IPDI) reacts with polycarbonate (PCDL) and polydimethylsiloxane (PDMS) polyol to produce a linear thermoplastic elastomer. Lignin is added to the matrix of the polyurethane, where lignin not only acts as a crosslinking agent but also as filler. The effects of polysiloxane-based polyurethane/lignin elastomers on thermal properties, mechanical properties, surface hydrophobicity, and cytotoxicity in vitro are investigated. The heat resistance, surface hydrophobicity and mechanical strength of the materials are improved, and a decreased cytotoxic behavior against HeLa cells by MTT assay is observed, which should be attributed to the semi-interpenetrating network structure and low toxicity of lignin, respectively. With the addition of 1 wt% of lignin, cell viability increases to 65.2%, mechanical strength increases to 17.1 Mpa and contact angle increases to 108.2°.
30 Jan 2020
TL;DR: Two isothermal section phase diagrams of the Zr-Ni-Sn ternary system at 973 K and 1173’K were experimentally constructed based on the binary phase diagrams, which provide an important instruction for the further optimization of ZrNiSn-based materials and other systems.
Abstract: The solubility range of interstitial Ni in the ZrNi Sn half-Heusler phase is a controversial issue, but it has an impact on the thermoelectric properties. In this study, two isothermal section phase diagrams of the Zr-Ni-Sn ternary system at 973 K and 1173 K were experimentally constructed based on the binary phase diagrams of Zr-Ni, Zr-Sn, and Ni-Sn. The thermodynamic equilibrium phases were obtained after a long time of heating treatment on the raw alloys prepared by levitation melting. Solubilities of at 973 K and at 1173 K were clearly indicated. An intermediate-Heusler phase with a partly filled Ni void was observed, which is believed to be beneficial to the lowered lattice thermal conductivity. The highest at 973 K was obtained for ZrNi1.11Sn1.04. The phase boundary mapping provides an important instruction for the further optimization of ZrNiSn-based materials and other systems.
Posted Content•
TL;DR: It is shown that the new elements lead to convergent algorithms for solving a gradcurl model problem as well as solving the Stokes system with precise divergence-free condition and the validity of the algorithms is demonstrated by numerical experiments.
Abstract: We construct finite element Stokes complexes on tetrahedral meshes in three-dimensional space. In the lowest order case, the finite elements in the complex have 4, 18, 16, and 1 degrees of freedom, respectively. As a consequence, we obtain gradcurl-conforming finite elements and inf-sup stable Stokes pairs on tetrahedra which fit into complexes. We show that the new elements lead to convergent algorithms for solving a gradcurl model problem as well as solving the Stokes system with precise divergence-free condition. We demonstrate the validity of the algorithms by numerical experiments.
TL;DR: In this paper, the authors used Ag to regulate the Zn-Sb framework for enhanced electrical properties, and prepared Eu2Zn1-0.5xAgxSb2 (x = 0.02, 0.06, and 0.1) by ball milling and spark plasma sintering.
TL;DR: A tiny amount of Mn is doped In0.15Sb1.85Te3 sample to tailor its carrier concentration, thus boosting the power factor and suppressing the bipolar effect, making this material a robust candidate for medium temperature (500~673 K) thermoelectric applications.
Abstract: A tiny amount of Mn is doped in In0.15Sb1.85Te3 sample to tailor its carrier concentration, thus boosting the power factor and suppressing the bipolar effect. Furthermore, large amounts of nanotwins are constructed to effectively scatter the phonons and reduce the lattice thermal conductivity. As a result, the zT value of Mn0.02In0.15Sb1.83Te3 is enhanced up to 1.0 at 673 K, making this material a robust candidate for medium-temperature (500-673 K) thermoelectric applications. Then combining with the low-temperature thermoelectric material Mn0.0075Bi0.5Sb1.4925Te3 previously reported by our group and using nickel as a barrier layer, a high average zT value of 1.08 during a broad temperature range from 303 to 673 K together with an Ohmic contact interface bonding is achieved in the p-type segmented leg fabricated via simple one-step sintering. Finally, the maximum theoretical conversion efficiency with a temperature difference of 370 K reaches ∼12.7%.
TL;DR: In this article, the authors used Yb to replace Ca to further tune thermoelectric transport behavior of CaMg1.1Bi1.98, an optimized composition according to previous work.
TL;DR: The results reveal that the LC is the tightest factor in regulating C-hydrolyzing enzyme activities, whereas the soil C quality (i.e. RIC) and flooding collectively drive C-Hydrolyzed enzyme activities possibly by affecting decomposition rates of SOC in the riparian zones.
Abstract: The effect of flooding on soil enzyme activities and soil organic carbon (SOC) dynamics remains a widely debated topic. Here, we investigated spatial variations in C-hydrolyzing enzyme activities, soil C contents in different fractions [i.e. labile and recalcitrant carbon (LC and RC)] from 6 sites with four different elevations at two soil depths (0–10 cm and 10–30 cm) in riparian zones of the Three Gorges Reservoir, China. At region scales, the SOC, RC contents, and RC/SOC (RIC) generally showed decreasing tendency from the upstream to the downstream. The C-hydrolyzing enzyme activities were higher in the midstream compared to other sites, which did not correspond well with the changing trend of SOC content, but matched with the spatial variation in LC content. At ecosystem scales, the RC and RIC declined with decreased elevations, but the LC showed opposite trend. Whereas, the four C-hydrolyzing enzyme activities and the specific enzyme activities were corresponded well with the changing trend of LC content. Soil C contents and enzyme activities were generally higher in top soil than deep soil across sites and elevation zones. These results reveal that the LC is the tightest factor in regulating C-hydrolyzing enzyme activities, whereas the soil C quality (i.e. RIC) and flooding collectively drive C-hydrolyzing enzyme activities possibly by affecting decomposition rates of SOC in the riparian zones.
Posted Content•
TL;DR: Three families of curl-curl conforming finite elements are constructed in two space dimensions, two of which contain two existing families and demonstrate the convergence and efficiency of the elements for solving the quad- curl problem.
Abstract: We construct smooth finite element de Rham complexes in two space dimensions. This leads to three families of curl-curl conforming finite elements, two of which contain two existing families. The simplest triangular and rectangular finite elements have only 6 and 8 degrees of freedom, respectively. Numerical experiments for each family demonstrate the convergence and efficiency of the elements for solving the quad-curl problem.
TL;DR: In this paper, a high performance and robust Sb2Te3-based alloys operating above 550 K is revealed, where nanograins and nanotwins are constructed by high energy ball milling and some boron particles are added to impede the boundary softening caused by the grain refinement.
TL;DR: In this paper, the substitution of the Bi site by isoelectric Sb was investigated and the effect of the Sb substitution on the transport behavior of AM2X2 compounds was investigated.
TL;DR: Bench scale experiments showed that different Fe2+ oxidation approaches worked well for phosphate removal at varied pH conditions, and Cu2+ exhibited superior catalytic Fe2+, Zn2+, and Ni2+.