Search

The distributed consensus mechanism is the backbone of the rapidly developing blockchain network. Blockchain platforms consume vast amounts of electricity based on the current consensus mechanism of Proof-of-Work (PoW). Here, we point out a different consensus mechanism named Proof-of-Stake (PoS) that can eliminate the extensive energy consumption of the current PoW-based blockchain. We comprehensively elucidate the current and projected energy consumption and carbon footprint of the PoW- and PoS-based Bitcoin and Ethereum blockchain platforms. The model of energy consumption of PoS-based Ethereum blockchain can lead the way toward the prediction of other PoS-based blockchain technologies in the future. With the widespread adoption of blockchain technology, if the current PoW mechanism continues to be employed, the carbon footprint of Bitcoin and Ethereum will push the global temperature above 1.5 °C in this century. However, a PoS-based blockchain can reduce the carbon footprint by 99% compared to the PoW mechanism. The small amount of carbon footprint from PoS-based blockchain could make blockchain an attractive technology in a carbon-constrained future. The study sheds light on the urgency of developing the PoS mechanism to solve the current sustainability problem of blockchain.

Relation Extraction (RE) is a crucial task in Information Extraction, which entails predicting relationships between entities within a given sentence. However, extending pre-trained RE models to other languages is challenging, particularly in real-world scenarios where Cross-Lingual Relation Extraction (XRE) is required. Despite recent advancements in Prompt-Learning, which involves transferring knowledge from Multilingual Pre-trained Language Models (PLMs) to diverse downstream tasks, there is limited research on the effective use of multilingual PLMs with prompts to improve XRE. In this paper, we present a novel XRE algorithm based on Prompt-Tuning, referred to as Prompt-XRE. To evaluate its effectiveness, we design and implement several prompt templates, including hard, soft, and hybrid prompts, and empirically test their performance on competitive multilingual PLMs, specifically mBART. Our extensive experiments, conducted on the low-resource ACE05 benchmark across multiple languages, demonstrate that our Prompt-XRE algorithm significantly outperforms both vanilla multilingual PLMs and other existing models, achieving state-of-the-art performance in XRE. To further show the generalization of our Prompt-XRE on larger data scales, we construct and release a new XRE dataset- WMT17-EnZh XRE, containing 0.9M English-Chinese pairs extracted from WMT 2017 parallel corpus. Experiments on WMT17-EnZh XRE also show the effectiveness of our Prompt-XRE against other competitive baselines. The code and newly constructed dataset are freely available at \url{https://github.com/HSU-CHIA-MING/Prompt-XRE}., IJCNN 2023

Based on the characteristic of doubly salient electromagnetic machines (DSEMs) that the incremental inductance varies with the rotor position, a sensorless startup strategy is proposed in this article. First, the pulse injection based method is adopted to estimate the initial rotor sector. The influence of the field current on the position estimation error is analyzed. Second, the startup method with no test pulse is introduced to start DSEM. Two key strategies are proposed: the commutation strategy based on the series incremental inductance detection and the anti-out-of-step control strategy based on the circulating current detection. The series incremental inductance is estimated in a chopping period and is compared with a preset threshold to obtain the commutation signal. The circulating current of the nonconducting phase is detected to monitor whether the DSEM is out of step. Furthermore, the current chopping mode is optimized, which has influence on the circulating current of the nonconducting phase. The proposed strategy can avoid the discontinuity of phase current caused by the test pulse in the existing technique. Finally, the effectiveness and the feasibility of the proposed method are verified by a series of experiments on the 9-kW 12/8 pole DSEM prototype.

Sulfur vacancy (SV) engineering is an evolving approach to improve the performance of metal sulfides for hydrogen evolution reactions (HER); however, a simple and universal method for creating SV i...

We describe the JD Explore Academy's submission of the WMT 2022 shared general translation task. We participated in all high-resource tracks and one medium-resource track, including Chinese-English, German-English, Czech-English, Russian-English, and Japanese-English. We push the limit of our previous work -- bidirectional training for translation by scaling up two main factors, i.e. language pairs and model sizes, namely the \textbf{Vega-MT} system. As for language pairs, we scale the "bidirectional" up to the "multidirectional" settings, covering all participating languages, to exploit the common knowledge across languages, and transfer them to the downstream bilingual tasks. As for model sizes, we scale the Transformer-Big up to the extremely large model that owns nearly 4.7 Billion parameters, to fully enhance the model capacity for our Vega-MT. Also, we adopt the data augmentation strategies, e.g. cycle translation for monolingual data, and bidirectional self-training for bilingual and monolingual data, to comprehensively exploit the bilingual and monolingual data. To adapt our Vega-MT to the general domain test set, generalization tuning is designed. Based on the official automatic scores of constrained systems, in terms of the sacreBLEU shown in Figure-1, we got the 1st place on {Zh-En (33.5), En-Zh (49.7), De-En (33.7), En-De (37.8), Cs-En (54.9), En-Cs (41.4) and En-Ru (32.7)}, 2nd place on {Ru-En (45.1) and Ja-En (25.6)}, and 3rd place on {En-Ja(41.5)}, respectively; W.R.T the COMET, we got the 1st place on {Zh-En (45.1), En-Zh (61.7), De-En (58.0), En-De (63.2), Cs-En (74.7), Ru-En (64.9), En-Ru (69.6) and En-Ja (65.1)}, 2nd place on {En-Cs (95.3) and Ja-En (40.6)}, respectively., WMT 2022 (Among all constrained systems, Vega-MT won 7 champions, 2 runners-up and 1 third place w.r.t sacreBLEU, and won 8 champions and 2 runners-up w.r.t COMET.)

BackgroundIntestinal ischemia/reperfusion (I/R) injury is a common tissue-organ damage occurring in surgical practice. This study aims to comprehensively review the collaboration and impact of countries, institutions, authors, subject areas, journals, keywords, and critical literature on intestinal I/R injury from a bibliometric perspective, and to assess the evolution of clustering of knowledge structures and identify hot trends and emerging topics.MethodsArticles and reviews related to intestinal I/R were retrieved through subject search from Web of Science Core Collection. Bibliometric analyses were conducted on Excel 365, CiteSpace, VOSviewer, and Bibliometrix (R-Tool of R-Studio).ResultsA total of 1069 articles and reviews were included from 2004 to 2022. The number of articles on intestinal I/R injury gradually plateaued, but the number of citations increased. These publications were mainly from 985 institutions in 46 countries, led by China and the United States. Liu Kx published the most papers, while Chiu Cj had the largest number of co-citations. Analysis of the journals with the most outputs showed that most journals focused on surgical sciences, cell biology, and immunology. Macroscopic sketch and microscopic characterization of the entire knowledge domain were achieved through co-citation analysis. The roles of cell death, exosomes, intestinal flora, and anesthetics in intestinal I/R injury are the current and developing research focuses. The keywords “dexmedetomidine”, “proliferation”, and “ferroptosis” may also become new trends and focus of future research.ConclusionThis study comprehensively reviews the research on intestinal I/R injury using bibliometric and visualization methods, and will help scholars better understand the dynamic evolution of intestinal I/R injury and provide directions for future research.

We propose an integrated path planning method for multiple automated guided vehicles performing logistics delivery within a real-world warehouse environment considering obstacles. By applying it on each vehicle, this proposed method enables the vehicle the vehicles have the capabilities for autonomous path planning. The path planning consists of three parts, K-means algorithm based task points clustering, genetic algorithm based task points ordering, and the probabilistic road map based best path search. Vehicle conflict resolution is depending on implementing the probabilistic road map construction considering the realistic map with obstacles. The simulations result validate that the clustering and ordering are necessary for the path planning, both the path planning time and the Automated guided vehicles (AGVs) running time can be dramatically reduced.

Background: As a kind of sliding bearing with liquid as a lubricant, Hydrostatic journal bearing can be shown from tribology, mechanics, optimization method and structural design in engineering practice. Hydrostatic bearing characterized by high precision, high rigidity, and long life has drawn people's attention and been widely adopted. As progress has been made in science and technology, people will have a better understanding of the principles and characteristics of the hydrostatic journal bearing. Objective: The development of hydrostatic bearing in recent years is studied and analyzed. Methods: More than 70 recently published patents and journal and conference papers related to the developments in air traffic flow management are reviewed, and the outcomes in temperature field, pressure field, deformation, lubrication, bearing capacity and friction are studied. Results: Through these research results, the current research status of hydrostatic bearings is summarized, and the three aspects of flow field characteristics, lubrication characteristics and tribology are explained, and solutions and feasibility measures for existing outstanding problems in each aspect are mentioned. Conclusion: Based on this article, the design and optimization of hydrostatic bearings are of guiding significance. Finally, the current problems and challenges of hydrostatic bearings are analyzed, and future research directions and research priorities are prospected.

For multilingual sequence-to-sequence pretrained language models (multilingual Seq2Seq PLMs), e.g. mBART, the self-supervised pretraining task is trained on a wide range of monolingual languages, e.g. 25 languages from CommonCrawl, while the downstream cross-lingual tasks generally progress on a bilingual language subset, e.g. English-German, making there exists the data discrepancy, namely domain discrepancy, and cross-lingual learning objective discrepancy, namely task discrepancy, between the pretraining and finetuning stages. To bridge the above cross-lingual domain and task gaps, we extend the vanilla pretrain-finetune pipeline with extra code-switching restore task. Specifically, the first stage employs the self-supervised code-switching restore task as a pretext task, allowing the multilingual Seq2Seq PLMs to acquire some in-domain alignment information. And for the second stage, we fine-tune the model on downstream data normally. Experiments on both NLG evaluation (12 bilingual translation tasks, 30 zero-shot translation tasks, and 2 cross-lingual summarization tasks) and NLU evaluation (7 cross-lingual natural language inference tasks) show our model outperforms the strong baseline mBART with standard finetuning strategy, consistently. Analyses indicate our approach could narrow the Euclidean distance of cross-lingual sentence representations, and improve the model generalization with trivial computational cost. We release the code at: https://github.com/zanchangtong/CSR4mBART.

Few-Shot classification aims at solving problems that only a few samples are available in the training process. Due to the lack of samples, researchers generally employ a set of training tasks from other domains to assist the target task, where the distribution between assistant tasks and the target task is usually different. To reduce the distribution gap, several lines of methods have been proposed, such as data augmentation and domain alignment. However, one common drawback of these algorithms is that they ignore the similarity task selection before training. The fundamental problem is to push the auxiliary tasks close to the target task. In this paper, we propose a novel task selecting algorithm, named Optimal Transport Task Selecting (OTTS), to construct a training set by selecting similar tasks for Few-Shot learning. Specifically, the OTTS measures the task similarity by calculating the optimal transport distance and completes the model training via a self-supervised strategy. By utilizing the selected tasks with OTTS, the training process of Few-Shot learning become more stable and effective. Other proposed methods including data augmentation and domain alignment can be used in the meantime with OTTS. We conduct extensive experiments on a variety of datasets, including MiniImageNet, CIFAR, CUB, Cars, and Places, to evaluate the effectiveness of OTTS. Experimental results validate that our OTTS outperforms the typical baselines, i.e., MAML, matchingnet, protonet, by a large margin (averagely 1.72\% accuracy improvement)., 12 pages, Under review of TKDE

This article aims to provide the information retrieval community with some reflections on recent advances in retrieval learning by analyzing the reproducibility of image-text retrieval models. Due to the increase of multimodal data over the last decade, image-text retrieval has steadily become a major research direction in the field of information retrieval. Numerous researchers train and evaluate image-text retrieval algorithms using benchmark datasets such as MS-COCO and Flickr30k. Research in the past has mostly focused on performance, with multiple state-of-the-art methodologies being suggested in a variety of ways. According to their assertions, these techniques provide improved modality interactions and hence more precise multimodal representations. In contrast to previous works, we focus on the reproducibility of the approaches and the examination of the elements that lead to improved performance by pretrained and nonpretrained models in retrieving images and text. To be more specific, we first examine the related reproducibility concerns and explain why our focus is on image-text retrieval tasks. Second, we systematically summarize the current paradigm of image-text retrieval models and the stated contributions of those approaches. Third, we analyze various aspects of the reproduction of pretrained and nonpretrained retrieval models. To complete this, we conducted ablation experiments and obtained some influencing factors that affect retrieval recall more than the improvement claimed in the original paper. Finally, we present some reflections and challenges that the retrieval community should consider in the future. Our source code is publicly available at https://github.com/WangFei-2019/Image-text-Retrieval., SIGIR 2022

Additional file 6: Table S2. Characteristics of patients with cardiopulmonary bypass.

Additional file 1: Table S1. Description of Cis-elements.

Additional file 4: Figure S3. Characterization of tissue injury and the gut microbiota in the Sen and Res mice. a The mRNA levels of proinflammatory factors in the ileum, liver, lung and kidney. b, c The total bacterial load and the level of Firmicutes and Bacteroidetes as well as the Firmicutes/Bacteroidetes ratio in the feces. d Relative bacteria abundance at the class level in the feces. e Heatmap of the expression values of 20 signaling pathways in each sample. f. Relative mRNA levels of ZO-1 and occludin in the ileum g Relative levels of ALT and AST in the plasma of GF mice. The results are expressed as the mean �� SEM. n = 6-8. ***p

Additional file 5: Figure S4. Lactobacillus murinus improves intestinal I/R-induced intestinal injury in antibiotic-treated mice. a. Experimental design. The WT mice were randomly divided into the following groups. Group ABX + I/R, in which ABX was administered 1w before intestinal ischemia; Group ABX + I/R + L. murinus, treatment of mice with ABX for a week, and then mice were gavaged daily for 7 days with L. murinus. b Relative abundance of L. murinus in the cecum after ABX clearance experiment. c-e HE staining and ZO-1 and occludin immunofluorescent staining in the ileum. Representative quantification on the right. Scale bar=100 ��m. f Tight junction mRNA levels in the ileum. g FD-4 level in the plasma. The results are expressed as the mean �� SEM. n = 8. ***p

Additional file 7: Figure S5. The level of IFABP and citrulline in serum before surgery between low L. murinus abundance group and high L. murinus abundance group.

Additional file 2: Figure S2. Experimental design. a L. murinus pretreatment experiment design. b Macrophage depletion experiment design.

Pre-Training (PT) of text representations has been successfully applied to low-resource Neural Machine Translation (NMT). However, it usually fails to achieve notable gains (sometimes, even worse) on resource-rich NMT on par with its Random-Initialization (RI) counterpart. We take the first step to investigate the complementarity between PT and RI in resource-rich scenarios via two probing analyses, and find that: 1) PT improves NOT the accuracy, but the generalization by achieving flatter loss landscapes than that of RI; 2) PT improves NOT the confidence of lexical choice, but the negative diversity by assigning smoother lexical probability distributions than that of RI. Based on these insights, we propose to combine their complementarities with a model fusion algorithm that utilizes optimal transport to align neurons between PT and RI. Experiments on two resource-rich translation benchmarks, WMT'17 English-Chinese (20M) and WMT'19 English-German (36M), show that PT and RI could be nicely complementary to each other, achieving substantial improvements considering both translation accuracy, generalization, and negative diversity. Probing tools and code are released at: https://github.com/zanchangtong/PTvsRI., Comment: COLING 2022

Additional file 8. Figure abstract.

Additional file 3: Table S1. Primers sequence.

Few-shot learning (FSL), purposing to resolve the problem of data-scarce, has attracted considerable attention in recent years. A popular FSL framework contains two phases: (i) the pre-train phase employs the base data to train a CNN-based feature extractor. (ii) the meta-test phase applies the frozen feature extractor to novel data (novel data has different categories from base data) and designs a classifier for recognition. To correct few-shot data distribution, researchers propose Semi-Supervised Few-Shot Learning (SSFSL) by introducing unlabeled data. Although SSFSL has been proved to achieve outstanding performances in the FSL community, there still exists a fundamental problem: the pre-trained feature extractor can not adapt to the novel data flawlessly due to the cross-category setting. Usually, large amounts of noises are introduced to the novel feature. We dub it as Feature-Extractor-Maladaptive (FEM) problem. To tackle FEM, we make two efforts in this paper. First, we propose a novel label prediction method, Isolated Graph Learning (IGL). IGL introduces the Laplacian operator to encode the raw data to graph space, which helps reduce the dependence on features when classifying, and then project graph representation to label space for prediction. The key point is that: IGL can weaken the negative influence of noise from the feature representation perspective, and is also flexible to independently complete training and testing procedures, which is suitable for SSFSL. Second, we propose Graph Co-Training (GCT) to tackle this challenge from a multi-modal fusion perspective by extending the proposed IGL to the co-training framework. GCT is a semi-supervised method that exploits the unlabeled samples with two modal features to crossly strengthen the IGL classifier.

Additional file 1: Figure. S1. Phylogenetic analysis of LGK ,AnmK and LGK-like proteins. Figure S2. Intracellular levoglucosan accumulation of CEN.PK113-5D and hxt-null strain EBY.VW4000. CEN.PK113-5D and EBY.VW4000 were cultured in YPD and YPM medium, respectively, for 12 h at 30 °C, then transferred into 20 mL fresh medium for another 12 h cultivation. The cells were then collected by centrifugation, washed twice with ddH2O, and resuspended in 30 mL YP medium supplemented with 5 gL−1 levoglucosan. After 30 min, 60 min, and 120 min incubation in 30 °C, respectively, cells were harvested and quickly washed with ice-cold ddH2O twice, then resuspended with ddH2O and placed at 37 °C overnight to extract the intracellular levoglucosan. The accumulation amounts were represented as mg levoglucosan per g dry cell weight (DCW). *, p

The distributed consensus mechanism is the backbone of the rapidly developing blockchain network. Blockchain platforms consume vast amounts of electricity based on the current consensus mechanism of Proof of Work. Here, we point out an advanced consensus mechanism named Proof of Stake that can eliminate the extensive energy consumption of the current PoW-based blockchain. We comprehensively elucidate the current and projected energy consumption and carbon footprint of the PoW and PoS based Bitcoin and Ethereum blockchain platforms.

With the application of more electric aircraft (MEA) technology, variable frequencies and high power ratings become import features of aero-generators. The brushless synchronous generator, which has a three-stage structure, is the most commonly used type of aero-generator. Due to the variation of operating conditions, the implementation of generator controllers becomes more and more difficult. In this paper, a state space model of a generator is derived and the influence of different operating conditions on the frequency response characteristics of the generator is revealed. Based on a fuzzy PI controller, an additional fuzzy logic controller is applied to modify the PI parameters of the voltage loop by introducing the generator speed to cope with the speed variation. Finally, the results of the simulations and experiments demonstrate that the dual fuzzy PI controller can improve both the steady-state and dynamic performance of the brushless synchronous generator, verifying the previous theoretical study.

A hybrid flame retardant, of excellent stability in aqueous media, was designed to develop flame-retarded polyethyleneimine (PEI) foam. In this case, iron phosphonate (FeP) with carboxyl group and hydroxyl group was first designed and synthesized. The carboxyl group could be hydrogen bonded with the water-soluble ammonium polyphosphate (APP) in aqueous media, and then dispersed as nano-sized particles. Subsequently, the well-dispersed nano-hybrid of FeP and APP (FeP/APP) was blended with hydrophilic PEI to prepare a kind of composite foam through a freeze-drying process. For FeP/APP, the transition metal iron showed excellent catalytic carbonization performances. Meanwhile, APP could also catalyze the degradation of polymers to form a protective char layer. The thermogravimetric analysis coupled with Fourier transform infrared analysis (TG-FTIR) test disclosed that the improvement of flame retardancy for PEI-based foams was ascribed to the synergistic effect of FeP and APP in the condensed phase. The composite foam containing 30 wt% FeP/APP could self-extinguish and reach V-1 rating in UL-94 test. When the loading level of 45% FeP/APP reached 45%, the composite foam could elevate up to V-0 rating, and the peak of heat release rate and total heat release were reduced by 71.8% and 74.2%, respectively, compared to a neat PEI foam. It is worth noting that our work presents a promising way for preparation of stable aqueous flame retardant, and is expected to enhance the fire safety of aqueous foams, coatings, cotton textiles, and other flammable materials.

Semi-solid forging of iron-based alloys during solidification has unique characteristics distinct from those of the classical hot forging. With the aim of acquiring precise knowledge concerning the microstructural evolution of bearing steel Cr4Mo4V in this process, a series of semi-solid forging experiments were carried out in which samples were wrapped in a designed pure iron sheath. The effects of forging temperature and forging reduction on the grain morphology and liquid flow behavior were investigated, respectively. By forging solidifying metal (FSM), bulky primary dendrites were broken and spheroidal grains with an average shape factor of 0.87 were obtained at 1360 °C. With the decreasing forging temperature to 1340 °C, the microstructural homogeneity can be improved. On the other hand, it shows that a higher forging reduction (50%) is essential for the spheroidization of grains and elimination of liquid segregation. Those microstructural characteristics are related to different motion mechanisms of solid and liquid phases at different forging temperatures. Additionally, the effect of semi-solid forging on the eutectic carbides was also investigated, and the results demonstrate that the higher diffusion capacity and less liquid segregation jointly lower the large eutectic carbides and consequently cause its uniform distribution during FSM.

Currently, deep learning (DL) algorithms have achieved great success in many applications including computer vision and natural language processing. Many different kinds of DL models have been reported, such as DeepWalk, LINE, diffusionconvolutional neural networks, graph convolutional networks (GCN), and so on. The GCN algorithm is a variant of convolutional neural network and achieves significant superiority by using a one-order localised spectral graph filter. However, only a one-order polynomial in the Laplacian of GCN has been approximated and implemented, which ignores undirect neighbour structure information. The lack of rich structure information reduces the performance of the neural networks in the graph structure data. In this study, the authors deduce and simplify the formula of two-order spectral graph convolutions to preserve rich local information. Furthermore, they build a layerwise GCN based on this two-order approximation, i.e. two-order GCN (TGCN) for semi-supervised classification. With the two-order polynomial in the Laplacian, the proposed TGCN model can assimilate abundant localised structure information of graph data and then boosts the classification significantly. To evaluate the proposed solution, extensive experiments are conducted on several popular datasets including the Citeseer, Cora, and PubMed dataset. Experimental results demonstrate that the proposed TGCN outperforms the state-of-art methods.

A sensorless control method based on the terminal voltage for doubly salient electro-magnetic motor (DSEM) is researched here. The negative-going zero-crossing point (NGZCP) of the fundamental component of terminal voltage is detected to obtain the rotor position. Firstly, a fourth Butterworth filter is introduced to extract the fundamental component of the terminal voltage. The ratio of signal frequency to cut-off frequency f/f c is optimised to acquire good filtering effect and good dynamic performance. Moreover, on this basis, the speed range is divided into several intervals to match the filters with different cut-off frequencies. Secondly, the coordinate transformation is adopted to compensate for the phase difference between the NGZCPs of the extracted voltages with the ideal commutation points. The transformation angle is obtained by offline simulation and test. An 18 kW DSEM prototype is tested to verify the feasibility of the proposed method.

To clarify the gel formation mechanism of isolated proteins recovered from blue round scads by isoelectric solubilization/precipitation, the physiochemical, conformational and gelation properties of the major protein myosin treated by acidic-(AM) or alkaline-(KM) pH-shifting were investigated and compared with native myosin (NM). The solubility in all myosin samples dropped significantly at 40 °C, with remarkable increases of particle size, turbidity and surface hydrophobicity. Among the samples, AM exhibited the lowest solubility and highest turbidity than NM and KM as the temperature was higher than 40 °C, indicating that AM is easily to be denatured and aggregated. Circular dichroism (CD) analysis showed that both acidic and alkaline pH treatments led to conformational changes of myosin, together with α-helix content variations. When pH was readjusted to neutral, KM recovered its structure similar to NM, while AM failed to refold to its native state. NM exhibited the highest storage modulus (G′) at 90 °C, while AM revealed the lowest G′ value. Furthermore, AM did not change obviously during heating, especially at the gelling temperature range (60–90 °C). These results strongly suggested that acidic treatment led to an irreversible conformational change in myosin, consequently induced a fierce denaturation and aggregation at the initial stage of heating, and eventually led to a complete loss of gel-forming ability. Alkaline pH-shifting, on the other hand is thus more feasible for protein recovery from blue round scads.

This article presents a sensorless control strategy based on the line-to-line excitation flux linkage for doubly salient electro-magnetic machine. The negative-going zero-crossing points of the line-to-line excitation flux linkage are detected to estimate the position and speed. To solve the problems of direct current (DC) bias and error accumulation in pure integrator, a combination of first-order low-pass filter and first-order high-pass filter is adopted to estimate the line-to-line flux linkage. Considering that the self-inductance obviously varies with the position and the phase current, an equivalent inductance is introduced to calculate the line-to-line excitation flux linkage in the whole period. The commutation error caused by the filter and the armature reaction is analysed and proved to be within ±6°. The proposed method can be applied to either the three-state standard angle control or the six-state advanced angle control. The experiment platform based on a 9-kW/3000-rpm DSEM is built, and the validity and feasibility of the proposed strategy are proved.

The titanium-based electrodes with MnOx nanoparticles coating (MnOx/Ti) and MnOx nanoparticles mixed with multi-walled carbon nanotubes (MnOx–CNTs/Ti) were fabricated by spraying–calcination method. The physicochemical properties of electrodes were investigated by SEM, XRD and XPS, which indicated that the surface coating of MnOx–CNTs/Ti, with MnOx nanoparticles dispersed uniformly on the CNTs, was smoother and with higher integrity than MnOx/Ti. Acid Red B was used as model pollutant to investigate the electro-catalytic activity of the electrodes, and the results revealed that the removal efficiency of Acid Red B reached 93.6% and 98.0% by MnOx/Ti and MnOx–CNTs/Ti, respectively, and the cell potential during the process of degradation by MnOx–CNTs/Ti was relatively low and stable. The electrochemical results confirmed that MnOx–CNTs/Ti possessed smaller charge transfer resistance and higher oxygen evolution current compared with MnOx/Ti, which can enhance the electro-catalytic activity and reduce the energy consumption by accelerating the transfer of electrons on the electrode surface. The accelerated lifetime tests of electrodes were carried out and showed that actual service lifetimes of MnOx–CNTs/Ti were 38 times of that for MnOx/Ti calculated by the experienced formula, which demonstrated that the durability of MnOx-based electrode was significantly promoted by addition of CNTs on Ti substrate.

Deep neural networks (DNNs) are known to be vulnerable to adversarial images, while their robustness in text classification is rarely studied. Several lines of text attack methods have been proposed in the literature, including character-level, word-level, and sentence-level attacks. However, it is still a challenge to minimize the number of word changes necessary to induce misclassification, while simultaneously ensuring lexical correctness, syntactic soundness, and semantic similarity. In this paper, we propose a Bigram and Unigram based adaptive Semantic Preservation Optimization (BU-SPO) method to examine the vulnerability of deep models. Our method has four major merits. Firstly, we propose to attack text documents not only at the unigram word level but also at the bigram level which better keeps semantics and avoids producing meaningless outputs. Secondly, we propose a hybrid method to replace the input words with options among both their synonyms candidates and sememe candidates, which greatly enriches the potential substitutions compared to only using synonyms. Thirdly, we design an optimization algorithm, i.e., Semantic Preservation Optimization (SPO), to determine the priority of word replacements, aiming to reduce the modification cost. Finally, we further improve the SPO with a semantic Filter (named SPOF) to find the adversarial example with the highest semantic similarity. We evaluate the effectiveness of our BU-SPO and BU-SPOF on IMDB, AG's News, and Yahoo! Answers text datasets by attacking four popular DNNs models. Results show that our methods achieve the highest attack success rates and semantics rates by changing the smallest number of words compared with existing methods., 12 pages, 3 figures, 10 tables

Biosynthesis of TCA cycle-derived C4 chemicals through glyoxylate shunt is an attractive metabolic route because it can be drived by TCA-glyoxylate cycle force under aerobic conditions. However, yield of this route is low with at least 1/3 carbon loss from glucose. FAs could sufficiently provide acetyl-CoA by β-oxidation without carbon loss and directly enter the TCA-glyoxylate cycle, which is acknowledged as a promising alternative feedstock. Here β-alanine was selected as the target TCA cycle-derived chemical, of which the theoretical yield is 1.391 g/g FAs, much higher than that of glucose(0.49 g/g). By adopting multi-metabolic engineering strategies and relieving the active oxygen damage caused by FAs utilization, β-alanine production reached 78.05 g/L with a yield of 1.2 g/g, about 86% of theoretical yield. Our study establish a promising bioproduction route of β-alanine from waste FAs (such as gutter oil, palm fatty acid distillate etc.), and more importantly, provide an efficient platform for TCA cycle-derived C4 chemicals biosynthesis.

In order to solve the problem that the accuracy of transformer condition assessment based on oil chromatographic data is not high, this paper adopts random forest to evaluate transformer state. In order to prevent the attribute values carried by the decision tree of the basic unit in the evaluation model from being too few and thus affecting the evaluation accuracy. The ratio of H2, CH4, C2H6, C2H4, C2H2, total hydrocarbon and various characteristic gases is used as the characteristic quantity, a stochastic forest optimization model for oil chromatography state assessment was constructedimmediately following the heading. It is compared with BP neural network and SVM, the test results of field oil chromatogram data show that the accuracy of random forest state assessment is 88.5%, which is better than BP and SVM, and the random forest assessment model has faster assessment speed, This paper provides technical support for the fast and accurate condition assessment of transformer.

Background. Osteosarcoma (OS) patients have a poor response to immunotherapy due to the sheer complexity of the immune system and the nuances of the tumor-immune microenvironment. Methodology. To gain insights into the immune heterogeneity of OS, we identified robust clusters of patients based on the immune gene expression profiles of OS patients in the TARGET database and assessed their reproducibility in an independent cohort collected from the GEO database. The association of comprehensive molecular characterization with reproducible immune subtypes was accessed with ANOVA. Furthermore, we visualized the distribution of individual patients in a tree structure by the graph structure learning-based dimensionality reduction algorithm. Results. We found that 87 OS samples can be divided into 5 immune subtypes, and each of them was associated with distinct clinical outcomes. The immune subtypes also demonstrated widely different patterns in tumor genetic aberrations, tumor-infiltrating, immune cell composition, and cytokine profiles. The immune landscape of OS uncovered the significant intracluster heterogeneity within each subtype and depicted a continuous immune spectrum across patients. Conclusion. The established five immune subtypes in our study suggested immune heterogeneity in OS patients and may provide optimal individual immunotherapy for patients exhibiting OS.

Visual attention in Visual Question Answering (VQA) targets at locating the right image regions regarding the answer prediction, offering a powerful technique to promote multi-modal understanding. However, recent studies have pointed out that the highlighted image regions from the visual attention are often irrelevant to the given question and answer, leading to model confusion for correct visual reasoning. To tackle this problem, existing methods mostly resort to aligning the visual attention weights with human attentions. Nevertheless, gathering such human data is laborious and expensive, making it burdensome to adapt well-developed models across datasets. To address this issue, in this paper, we devise a novel visual attention regularization approach, namely AttReg, for better visual grounding in VQA. Specifically, AttReg firstly identifies the image regions which are essential for question answering yet unexpectedly ignored (i.e., assigned with low attention weights) by the backbone model. And then a mask-guided learning scheme is leveraged to regularize the visual attention to focus more on these ignored key regions. The proposed method is very flexible and model-agnostic, which can be integrated into most visual attention-based VQA models and require no human attention supervision. Extensive experiments over three benchmark datasets, i.e., VQA-CP v2, VQA-CP v1, and VQA v2, have been conducted to evaluate the effectiveness of AttReg. As a by-product, when incorporating AttReg into the strong baseline LMH, our approach can achieve a new state-of-the-art accuracy of 60.00% with an absolute performance gain of 7.01% on the VQA-CP v2 benchmark dataset..., ACM TOMM 2021

Additional file 2: Table S1. The primers used in this study.

Additional file 1: Figure S1. The relative GFP intensity of the Pfrd3 promoter at different OD conditions. Figure S2. OD optimization and rotation speed optimization. (A) Production comparison between different OD conditions. (B) Production comparison between different rotation speeds. Figure S3. Remaining soybean oil of Fed-batch fermentation in 1-L bioreactors for RK production.

Additional file 2: Table S1. The primers used in this study.

Additional file 1: Figure S1. Identification of germacrene A by GC-MS analysis. (A, B) GC analyses of the ß-elemene standard (A) and the culture extract of the yeast strain SGAS5. (C) GC-MS spectra of the chromatographic peak corresponding to ß-elemene in (B). Figure S2. List of pairwise protein sequence identity (%) comparison between ten GAS candidates applied for screening in this study. Figure S3. Extracellular farnesol production of yeast strains SGAS5, SMVA1, SMVA2, and SMVA3. Figure S4. Sequence alignment of AvGAS with NsGAS, NcGAS, and bLinS from Streptomyces clavuligerus (5NX6). Two conserved regions within the active site, the aspartate-rich (DDXX(X)(D,E)) motif and the NSE (NDXXSXX(R,K)(E,D)) triad, required for binding catalytically essential Mg2+ ions, were indicated with boxes. Figure S5. Extracellular germacrene A production of yeast cells expressing the indicated GASs, respectively. GASs were expressed in S. cerevisiae CEN.PK2-1D wherein tHMGR1 was overexpressed and ERG9 was downregulated via promoter replacement, as described in the main text. Figure S6. Extracellular farnesol production of yeast strain SHEF and SHEF strains expressing LsGAS, AvGAS, and AvGAS-F23W, respectively. Figure S7. Extracellular and intracellular germacrene A accumulation in the yeast strain SHEF after cultivation for 72 h. Figure S8. Location of Phe23 in the modelled structure of AvGAS. The substrate FPP is represented as cyan sticks and Phe23 is shown as purple sticks. Table S1. Growth analyses of yeast cells expressing Phe23 mutants of AvGAS after cultivation for 72 h.

Additional file 1: Table S1. Primers used in this study. Table S2. Amino acid sequence identities (%) between α-1, 2-FTs tested in the present study and selected characterized α-1, 2-FTs in literatures. Figure S1. Lactose importing analysis by strain FL01 and FL02. Figure S2. Phylogenetic analysis and alignment of α-1, 2-FTs of different organisms including α-1, 2-FTs tested in the present study and selected α-1, 2-FTs reported in the literatures. Figure S3. Western blot assay of His-tagged FutBc and FutC expressed in yeast cells. Figure S4. 2′-FL production by FL06 with different initial lactose concentration. Figure S5. Cell wall stability test of strains producing 2′-FL. Figure S6. Effect of overexpression Pmi40 and sPos5 on 2′-FL production.

This paper focus on the state estimation, number of targets estimation and target type estimation of mixed targets and propose a mixed target tracking algorithm with type probability using the GLMB filter. The mixed targets include point target, extended target and group target (indistinguishable). The tracking algorithm mainly includes three aspects: mixed target dynamic modeling, type analysis and tracking estimation. Firstly, measurement model of mixed target is established with the generalized label multi-Bernoulli filter. Secondly, the probability of targets type are calculated. Finally, the parameters of finite mixture models (FMM) are derived by using Gibbs sampling and BIC criterion to track the mixed target. Then the equivalent measurement method is used to replace the measurement of extended target and group target. The shape of the mixed target is estimated by the ellipse approximation modeling method. Simulation experiments are provided to show the effectively of track mixed targets.

Real world traffic sign recognition is an important step towards building autonomous vehicles, most of which highly dependent on Deep Neural Networks (DNNs). Recent studies demonstrated that DNNs are surprisingly susceptible to adversarial examples. Many attack methods have been proposed to understand and generate adversarial examples, such as gradient based attack, score based attack, decision based attack, and transfer based attacks. However, most of these algorithms are ineffective in real-world road sign attack, because (1) iteratively learning perturbations for each frame is not realistic for a fast moving car and (2) most optimization algorithms traverse all pixels equally without considering their diverse contribution. To alleviate these problems, this paper proposes the targeted attention attack (TAA) method for real world road sign attack. Specifically, we have made the following contributions: (1) we leverage the soft attention map to highlight those important pixels and skip those zero-contributed areas - this also helps to generate natural perturbations, (2) we design an efficient universal attack that optimizes a single perturbation/noise based on a set of training images under the guidance of the pre-trained attention map, (3) we design a simple objective function that can be easily optimized, (4) we evaluate the effectiveness of TAA on real world data sets. Experimental results validate that the TAA method improves the attack successful rate (nearly 10%) and reduces the perturbation loss (about a quarter) compared with the popular RP2 method. Additionally, our TAA also provides good properties, e.g., transferability and generalization capability. We provide code and data to ensure the reproducibility: https://github.com/AdvAttack/RoadSignAttack., IEEE Internet of Things Journal 8 (6), pp. 4980-4990, 2021

Domain adaptation aims to generalize a model from a source domain to tackle tasks in a related but different target domain. Traditional domain adaptation algorithms assume that enough labeled data, which are treated as the prior knowledge are available in the source domain. However, these algorithms will be infeasible when only a few labeled data exist in the source domain, and thus the performance decreases significantly. To address this challenge, we propose a Domain-invariant Graph Learning (DGL) approach for domain adaptation with only a few labeled source samples. Firstly, DGL introduces the Nystrom method to construct a plastic graph that shares similar geometric property as the target domain. And then, DGL flexibly employs the Nystrom approximation error to measure the divergence between plastic graph and source graph to formalize the distribution mismatch from the geometric perspective. Through minimizing the approximation error, DGL learns a domain-invariant geometric graph to bridge source and target domains. Finally, we integrate the learned domain-invariant graph with the semi-supervised learning and further propose an adaptive semi-supervised model to handle the cross-domain problems. The results of extensive experiments on popular datasets verify the superiority of DGL, especially when only a few labeled source samples are available.

Numerous studies have reported that hydroxyapatite nanoparticles (nano-HAPs) can inhibit the proliferation of a variety of tumor cells and this effect is different in different carcinoma cells. However, the effect of nano-HAPs on osteosarcoma cell proliferation has not been well understood thus far. In this study, we first showed that our synthesized nano-HAPs reduced cell viability and inhibited migration and invasion of OS-732 cells in a concentration-dependent manner. Using a BALB/c nude mouse tumor model, we demonstrated that nano-HAPs could effectively suppress tumor growth in vivo. We also performed RNA-seq analysis to investigate the underlying mechanism of these effects and discovered that treatment of OS-732 cells with nano-HAPs significantly downregulated the FAK/PI3K/Akt signaling pathway. Collectively, our study suggests that treatment with nano-HAPs can inhibit osteosarcoma cell growth, migration and invasion in vitro and suppress osteosarcoma in vivo.

The contact mechanics model of the metal lens-type sealing gasket is established on the basis of Hertz theory on the macroscopical scale in this paper. The relationship among sealing width, contact pressure, and preload is solved. Based on the structural characteristics of the subsea collet connector, the self-locking characteristics are analyzed to determine the gain coefficient of the sealing structure for the loading thrust. On the microscopic scale, the contact characteristics of the turning lens-type sealing gasket and the hub structure are analyzed by the equivalent replacement of the peak cut coefficient of the one-dimensional sinusoidal wave. The influence of different leakage forms on sealing performance is discussed from both radial and circumferential leakage, and the leakage rate of the lens-type sealing structure is calculated. The hydrostatic pressure experiment of the subsea collet connector with lens-type sealing gasket is carried out, and the correctness of the theoretical analysis is verified from the results of the pressure maintaining, sealing width measurement, and preload conversion.