Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2024-01-04
Abstract： Neural Transducer and connectionist temporal classification (CTC) are popular end-to-end automatic speech recognition systems. Due to their frame-synchronous design, blank symbols are introduced to address the length mismatch between acoustic frames and output tokens, which might bring redundant computation. Previous studies managed to accelerate the training and inference of neural Transducers by discarding frames based on the blank symbols predicted by a co-trained CTC. However, there is no guarantee that the co-trained CTC can maximize the ratio of blank symbols. This paper proposes two novel regularization methods to explicitly encourage more blanks by constraining the self-loop of non-blank symbols in the CTC. Experiments on LibriSpeech corpus show that our proposed method accelerates the inference of neural Transducer by 4 times without sacrificing performance. It is interesting to find that the frame reduction ratio of the neural Transducer can approach the theoretical boundary. Additionally, a further gain can be observed when decoding with external language models. To the best of our knowledge, this is the first work to explore the feasibility of neural Transducers with almost no blank symbols.
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2021-11-25
Abstract： Driver is the core of the "human-vehicle-environment" road traffic system. It is of great significance to study the perception-action pattern of drivers for standardizing driving behavior and improving safety level. However, there are few studies on head movement and organ cooperation during driving, especially on quantitative calculation. Therefore, this paper designed an experiment with the goal of uniform-speed driving in the virtual reality environment, and used information theory tools for modeling and analysis. We studied the cooperation of head, eye, hand and foot, proposed a causal network based on transfer entropy to describe the cooperation mode between the four, and proposed to use the network average transfer entropy as an indicator to evaluate the coordination of organs in driving. Finally, we found that head-eye cooperation is very strong in driving. The cooperation between organs is better when turning than when going straight.In the uniform-speed driving task, the priority of the driver to the action task is higher than the priority of the perception task."
Peer Review Status:
Awaiting Review
Subjects: Computer Science >> Other Disciplines of Computer Science submitted time 2021-10-21
Abstract： Active information storage is an important growing popularity information theoretic tool, which has the advantage of being an easily accessible and interpretable store of information that is complex in the system. Driver head movement behaviour plays an important role in their directional control, yet this role has not been measured and explained quantitatively. In this paper, we apply active information storage to the study of driver head movements to investigate their relationship with driving directional smoothness. Specifically, we design a VR driving experiment with containing straight and turning directions, obtain a sequence of active driver head movement information storage and a sequence of vehicle deflection angles, and investigate the quantitative relationship between the two. We demonstrate a high temporal correlation between the two and use the joint entropy of the two as an indicator of driving performance. Finally, we use driver head movements to predict vehicle deflection angles in real time and obtain an accuracy of 88.56%. This work is expected to help monitor driver state and improve driving safety.
Peer Review Status:Awaiting Review
Subjects: Hydraulic Engineering >> Basic Disciplines of Hydraulic Engineering submitted time 2021-01-30
Abstract： From the perspective of systematic thinking, this study applied the theories of information entropy and dissipative structure to comprehensively evaluate and analyze the evolution of the complex system of the Yellow River Basin over the past 40 years. With the concept of “Basin Development Index (BDI shortly)”, a preliminary scientific comprehensive evaluation method system of basin development quality is established. Results of the model calculation show that the ratio of per capita GDP in the basin to the national average and the total sedimentation in the lower Yellow River have occupied the largest weight for many years, and have had a relatively greater impact on the development of the basin in different historical periods. During the entire 40 years, the value of BDI has varied from 50.5 points in 1996 to 66.6 points in 2019, and has generally experienced a decreasing and an increasing stage. The evolution of BDI of Yellow River basin is consistent with the overall law of the basin development and has a great correlation with some major events happened in the basin. In addition, this research method can be extended to other fields such as regional governance. Compared with the current evaluation models mainly based on economic indicators, this novel methodology is more systematic and scientific, and will better serve the decision-making with more scientific thoughts.
Peer Review Status:Awaiting Review
Subjects: Hydraulic Engineering >> River Silt Engineering submitted time 2021-01-29
Abstract： This article systematically summarized the development process of the operation mode of reservoirs on sediment-laden rivers in China since the 1950s, especially explained the application and design technology of "storing clean water and regulating muddy flow". The so-called " storing clean water and regulating muddy flow " application is to fully consider the changes in the incoming water and sediment process in the reservoir according to the requirements of the reservoir development task, and coordinate the impact of the sediment on the siltation form and effective reservoir capacity to maximize the coordination degree of the water-sediment relationship in the downstream river as the core. Through the comprehensive coordinated control of "blocking, regulating and discharging", the comprehensive utilization benefits of reservoirs can be brought. In design, in order to better exert the reservoir regulation function, it is necessary to consider the demand for storage capacity of water and sediment regulation, and set up sufficient water and sediment regulation storage capacity to reduce the adverse impact of the forced sediment discharge of the reservoir on the downstream water-sediment relationship. The storage capacity distribution design is based on the rules of “deep channel for sediment regulation, medium channel for benefit and high channel for flood regulation”, to avoid the use of effective reservoir capacity due to sedimentation. Rivers with ultra-high sediment content should set abnormal sediment discharge bottom outlets to form a "normal + abnormal" double sediment erosion datum plane, to realize the recycling of part of the sediment storage capacity, and the integrated utilization of sediment storage capacity and water and sediment regulation storage capacity. Reservoir should adopt water and sediment separation development method, which is “"mainstream large reservoirs for sediment regulation, tributary small reservoirs for water supply", in rivers with super high sediment content to solve the contradiction between effective reservoir capacity maintenance and water supply regulation. On the basis of achieving long-term preservation of effective storage capacity, "storing clean water and regulating muddy flow" better takes into account the needs of reservoir water and sediment regulation in sediment-laden rivers. The results will provide guidance and reference for the reservoir engineering design and operation of sediment-laden river, in the current and future periods.
Peer Review Status:Awaiting Review
Subjects: Information Science and Systems Science >> Basic Disciplines of Information Science and Systems Science submitted time 2021-01-26
Abstract： It is an important decision-making basis for the high-quality development of the whole basin in mastering the evolution of the spatio-temporal pattern in the economic development of the Yellow River Basin systematically. At present, GDP is the main reference to measure the pattern of economic development, however, GDP statistics have the disadvantages of not uniform caliber and low spatial resolution, it is difficult to accurately characterize the evolution pattern of economic development in space and time. The nightlight data reflect it effectively with its relatively objective characteristics. Through the correction and processing of the national nightlight data from 1992 to 2013, this paper obtains the national and watershed nightlight data images, conducts economic spatial aggregation analysis, and compares and analyzes the relevant factors. The results show that the areas with high economic spatial aggregation are distributed near the Yellow River channel, from the upper reaches to the lower reaches of the Yellow River, the heat intensity is gradually increasing, and the provincial capital city is the core area, spreading outwards, showing the tendency of the polymerization area to disperse to the surrounding area under the condition that the polymerization of the old polymerization point is intensified; The basin development index BDI has a certain correlation, the overall trend of nightlight data in 9 provinces of the basin is basically consistent with the change trend of the national nightlight data, showing a steady upward trend, but the overall value is lower than the national nightlight data, the high-quality development of the Yellow River basin is imminent.
Peer Review Status:Awaiting Review
Subjects: Psychology >> Other Disciplines of Psychology submitted time 2020-04-22
Abstract： How to optimize entrepreneurial behavior and thus improve the quality of entrepreneurship is of central interest for both academics and practitioners. As a salient characteristic of entrepreneurial behavior is its dynamic match with external environment, research attention shifting from a focus on static entrepreneurial behavior to tracking dynamic changes in entrepreneurial behavior is warranted. In this study, we posit that behavior is simultaneously influenced by personality, cognition, and environment, and explore how an entrepreneur’s narcissism affects his/her behavioral change and following performance. Further, we investigate how narcissism interacts with experience in affecting behavioral change and its performance implications across different cognitive contexts and business environments. Our findings contribute to the research on entrepreneurial behavior and entrepreneurial psychology, as well as accumulate knowledge on how to help entrepreneurs understand themselves and improve their ability of decision making.
Peer Review Status:Awaiting Review
Subjects: Biology >> Bioengineering submitted time 2019-05-06
Abstract： " Terpenoids have considerable commercial value, but the production process is complex and the yield is low. It has become a hot spot to synthesize terpenoids from microorganism. Corynebacterium glutamicum contains a pathway to produce carotenoid, which is a natural advantage for the synthesis of terpenoids heterologously. The synthesis of terpenoids from C. glutamicum is summarized, including the terpenoids synthesis pathway in C. glutamicum, key enzymes and global regulatory mechanisms in this pathway. And the advances in this pathway in synthesis of monoterpenes, sesquiterpenes, and tetraterpenes are summarized. The problems and advice efficient synthesis of terpenoids by C. glutamicum is discussed.
Peer Review Status:Awaiting Review
Subjects: Biology >> Bioengineering submitted time 2017-07-24 Cooperative journals: 《中国生物工程杂志》
Abstract： CRISPR/Cas9 system: A recent progress
Subjects: Biology >> Bioengineering submitted time 2017-07-24 Cooperative journals: 《中国生物工程杂志》
Abstract：8二甲基异戊烯基柚皮素（8DN）作为生产黄酮类类药物淫羊藿苷的重要前体，在医药合成领域具有重大应用潜力。由于其合成路径及相关基因的复杂性，目前主要通过饲喂8DN的直接前体(柚皮素，异黄腐酚等）的方式合成8DN，而在生物体内全合成8DN的的研究工作还未见报道。为了实现8DN在酿酒酵母体内的生物全合成，该研究通过组合筛选8DN前体物柚皮素合成所需的多种外源基因（TAL，4CL，CHS，CHI），获得30株柚皮素生产菌，发现不同来源的基因组合引起柚皮素产量的明显差异（0.37mg/L-22.33mg/L）。并且利用Delta位点将较优的基因组合整合至酵母基因组，实现了稳定的柚皮素高产菌株（SyBE_Sc02050031）构建。在此基础上进一步导入带有苦参来源的异戊烯基转移酶基因（N8DT）多拷贝质粒，实现8DN合成的完整反应过程，8DN的摇瓶发酵产量达到36.7μg/L。另外通过关键限速酶N8DT的序列优化策略，发现截断定位信号肽序列的N8DT明显提高了从柚皮素到8DN这一关键反应的催化效果，8DN的产量提高到52.6 μg/L（144.2%）。该研究首次在酿酒酵母中成功构建高产8DN的生物全合成路径，为在微生物体内合成其他黄酮类天然产物提供了参考，具有重要的指导意义。
Subjects: Biology >> Bioengineering submitted time 2017-07-24 Cooperative journals: 《中国生物工程杂志》
Abstract：9β,11β-环氧-17α,21-二羟基-16β-甲基孕-1,4-二烯-3,20-二酮（Ⅳ）是生产9-氟甾体激素的关键前体，以9β,11β-环氧-17α,21-二羟基-16β-甲基孕-4-烯-3,20-二酮-21-醋酸酯（Ⅰ）为底物合成Ⅳ是工业化生产Ⅳ的重要方法。通过比较分枝杆菌全细胞转化法与细胞裂解液转化法，发现分枝杆菌全细胞只能将Ⅰ转化9β,11β-环氧-17α,21-二羟基-16β-甲基孕-4-烯-3,20-二酮（Ⅱ），而细胞裂解液可以有效地将Ⅰ转化为Ⅳ，其反应机制为底物Ⅰ自发水解为中间体Ⅱ，Ⅱ在C1,2位脱氢酶（KSTD）的催化作用下发生C1,2位脱氢反应生成产物Ⅳ。为进一步提高产物Ⅳ的生成率，利用基因工程手段在分枝杆菌中分别过表达编码KSTD的关键基因：kstD、kstD3和kstDM，提高脱氢反应效率，结果表明1 g/L底物Ⅰ在pH 7.0的重组菌株MS136-kstDM细胞裂解液中反应45 h，Ⅳ的生成率为78.4 %，比出发菌株提高了38.9％；并优化缓冲液pH，提高反应速率，结果表明1 g/L底物Ⅰ在pH 7.5的重组菌株MS136- kstDM细胞裂解液中反应45 h，Ⅳ的生成率为92.8 %，比出发菌株提高了63.4 %。
Subjects: Biology >> Bioengineering submitted time 2017-07-24 Cooperative journals: 《中国生物工程杂志》
Abstract：巴氏芽孢八叠球菌是迄今为止所知的利用尿素降解进行生物矿化的最为高效的微生物系统之一，基于其碳酸钙生物矿化形成的“超强能力”，巴氏芽孢八叠球菌已经被成功地应用于沙石、土壤等生物固化中，成为一种全新的、具有极大潜力的生物建筑辅助技术，被称为“生物水泥”。由于巴氏芽孢八叠球菌分离自土壤，没有病原性并具有极好的环境友好性。近年来，其应用领域被拓展到环境治理乃至健康医疗领域，成为全新的研究热点。然而，与应用研究相比，人们对巴氏芽孢八叠球菌生物矿化背后的分子机理还知之甚少。因此，这里针对迄今为止国内外对于巴氏芽孢八叠球菌生物矿化相关的分子机理的研究进行分析介绍，在此基础上综述了在包括建工、环境、及医疗健康领域内巴氏芽孢八叠球菌的应用，希望能够促进对于巴氏芽孢八叠球菌矿化的研究。
Subjects: Materials Science >> Materials Science (General) submitted time 2016-11-04 Cooperative journals: 《金属学报》
Abstract：摘 要 利用扫描电镜（SEM）和透射电镜（TEM）研究了Inconel625焊丝熔敷金属中δ相的形成机理，并揭示了δ相粗化的内在规律。结果表明，热处理后熔敷金属中大量析出的针状δ相呈网格状分布在基体 相中，且其附近出现了贫 相区域。δ相的形成是一个类贝氏体转变的固态相变过程，其晶核是在 相密排面的层错上通过切变方式形成的，同时δ相的粗化行为是扩散控制的生长过程。保温时间较短时，实际δ相颗粒的平均尺寸符合LSW理论，随着保温时间的延长，δ相的高密度、不定向析出特征使得其实际尺寸偏离了经典LSW理论的预测值。
Hits
YES
Hits
Downloads
Comment