Subjects: Information Science and Systems Science >> Information Security Technology submitted time 2023-11-23
Abstract： In the online ride-hailing platform, the order dispatch algorithm serves as the infrastructure for providing services to match drivers and passengers, which directly affects the rights and travel experience of drivers and passengers. How to test and ensure the fairness of the order dispatch algorithm is critical to the orderly operation of the online ride-hailing platform and protecting the rights and interests of both drivers and passengers. This paper proposes a non-intrusive fairness test method for online ride-hailing order dispatch algorithms. There is no need to go into the details of the algorithm. Only the scene information and the dispatch results are needed to conduct the algorithm fairness test. This method greatly reduces the calculation cost and can protect corporate confidentiality and user privacy. It first characterizes algorithm fairness through statistical modeling of randomness. Then it uses hypothesis testing methods to verify algorithm fairness, and finally uses zero-knowledge proof technology to build contactless trust. In the study, Didi Chuxing data was used to verify the fairness of the algorithm based on pick-up distance. The experiment verified the feasibility and effectiveness of this method, and demonstrated the potential of this framework for larger-scale and more complete algorithm fairness testing.
Peer Review Status:
Awaiting Review
Subjects: Computer Science >> Natural Language Understanding and Machine Translation Subjects: Library Science，Information Science >> Machine Translation submitted time 2023-04-18
Abstract：目前主流的人工智能，普遍采用注意力机制 + 深度学习+强化学习的技术道路。我们认为强化学习无法适用到那些难以大量试错的领域。所以，要想实现能适用于任何领域的通用人工智能，我们必须转变实现道路。所以，我们提出了一套不同于深度学习+强化学习的机器学习方案，它通过小样本、累积学习，同样实现了和 transformer 相似的注意力机制，也同样创建了全连接知识网络。并且，它不需要采用试错学习的方式，就可以实现和环境的互动决策。并且人类可以给它预置不同的先天需求，来实现多目标平衡，从而实现远高于目前人工智能的安全性。在本文中，我们提出了一套从0 到1 的新机器学习技术方案。
Peer Review Status:Awaiting Review
Subjects: Other Disciplines >> Synthetic discipline submitted time 2023-03-28 Cooperative journals: 《中国科学院院刊》
Abstract： At the dawn of the 4th Industrial Revolution, cultivating top innovation talents that innovate “zero-to-one” achievements, and constructing sustainable mechanism for cultivating such talents, is not only a critical task during the China’s national rejuvenation in the world experiencing a level of change unseen in a century, but also the key parameter for centenary or even millennium prosperity of Chinese nation. After over ten years of exploration, Tsien Excellence in Education Program of Tsinghua University (hereafter called “Tsien Class”) has made a single-point breakthrough in this subject. Based on this, nearly after one year of preparation, “Tsien Class” and Shenzhen government jointly founded the Xinstitute of Shenzhen, an institution with the mission of cultivating top innovation talents, which embodies the institutional and cultural advantages of China. We believe that, following this direction for another ten years, we will be able to establish a system, which is initiated in Shenzhen and radiates the whole country, that provides a new paradigm to Higher Education to cultivate top innovation talents.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2020-09-28 Cooperative journals: 《计算机应用研究》
Abstract： 3d seismic acoustic wave theory and calculation method are the basis of geological exploration research. By analyzing the propagation characteristics of acoustic waves in different media, we can apply the 3d seismic acoustic wave forward mod-eling in exploration work. In order to solve the problem of huge numerical calculation and large memory consumption while proceeding 3D seismic wave equation staggered grid finite difference forwarding model, we studied and implemented the parallel optimization on the heterogeneous many-core processors of the Sunway Taihulight supercomputer. Based on the implementation of a two-level parallel programming model by using MPI+Sunway Athread, we generated the DMA communication, 2.5D pipelining task division and other optimization strategies. The model with this improvement reduced the negative effects by bandwidth and greatly utilized the computing power. In large-scale conditions, we tackled the issue of low efficiency about program execution on SW26010 het-erogeneous many-core processors. The experimental results reveal that the performance of parallel-ism of a single node is much better than that of a master core. Another example is the calculation of solving numeri-cal stress is 80 times faster on the core group than that on the single master core. This experiment could keep a con-stant performance by using 266240 cores, 5.5GFlpos updates of field.
Subjects: Psychology >> Management Psychology submitted time 2020-04-02
Abstract： The innovation process can be conceptualized as encompassing two stages: idea generation and idea implementation. However, to date, more early research has been focused on the idea generation stage, and these two subfields still remain stubbornly disconnected from one another. To integrate these two disparate subfields and advance our comprehensive understanding of the innovation process, the current study relied on self-determination theory and Yale attitude change approach to examine the effect of intrinsic motivational orientation of idea generators on idea quality, and further tested the effect of idea quality, rewards of the idea generators, and their interaction on idea implementation. To test the hypotheses in the proposed model, we collected data from 251 idea generators, 7 idea evaluators and the manager of an innovative proposal project at 3 different time points in a large financial company. In the first wave, 440 ideas proposed by 440 idea generators were recorded. In the second wave, all of the idea generators were invited to complete a questionnaire, including demographic information and intrinsic motivational orientation, and the number of valid questionnaires was 251; moreover, 7 in-house experts assessed the idea generators and the quality of each idea. Approximately six months later, the manager of this project reported the implementation degree of all of the ideas. Our hypotheses were examined using SPSS 22.0. By analyzing the multi-time and multi-source data, it was found in this study that: (1) intrinsic motivational orientation of the idea generators has a significantly positive effect on idea quality; (2) idea quality further facilitates idea implementation; (3) rewards of the idea generators positively influence idea implementation; and (4) idea quality and rewards of the idea generators interact to predict idea implementation.The findings presented here contribute to innovation literature in the following ways. First, this study tracks the innovation process from idea generation to idea implementation, bridging the theoretical divide between creativity research and idea implementation research. Second, this study clarifies the positive effect of intrinsic motivational orientation on idea quality, which is a beneficial supplement to, and promotion of the existing research on the relationship between intrinsic motivation and creativity. Third, Yale attitude change theory is introduced into idea implementation research, which not only expands the application scope of this theory, but also enriches theoretical perspectives of idea implementation research.
Subjects: Psychology >> Management Psychology submitted time 2020-02-25
Abstract： The innovation process can be conceptualized as encompassing two stages: idea generation and idea implementation. However, to date, more early research has been focused on the idea generation stage, and these two subfields still remain stubbornly disconnected from one another. To integrate these two disparate subfields and advance our comprehensive understanding of the innovation process, the current study relied on self-determination theory and Yale attitude change approach to examine the effect of intrinsic motivational orientation of idea generators on idea quality, and further tested the effect of idea quality, rewards of the idea generators, and their interaction on idea implementation. To test the hypotheses in the proposed model, we collected data from 251 idea generators, 7 idea evaluators and the manager of an innovative proposal project at 3 different time points in a large financial company. In the first wave, 440 ideas proposed by 440 idea generators were recorded. In the second wave, all of the idea generators were invited to complete a questionnaire, including demographic information and intrinsic motivational orientation, and the number of valid questionnaires was 251; moreover, 7 in-house experts assessed the idea generators and the quality of each idea. Approximately six months later, the manager of this project reported the implementation degree of all of the ideas. Our hypotheses were examined using SPSS 22.0. By analyzing the multi-time and multi-source data, it was found in this study that: (1) intrinsic motivational orientation of the idea generators has a significantly positive effect on idea quality; (2) idea quality further facilitates idea implementation; (3) rewards of the idea generators positively influences idea implementation; and (4) idea quality and rewards of the idea generators interact to predict idea implementation. The findings presented here contribute to innovation literature in the following ways. First, this study tracks the innovation process from idea generation to idea implementation, bridging the theoretical divide between creativity research and idea implementation research. Second, this study clarifies the positive effect of intrinsic motivational orientation on idea quality, which is a beneficial supplement to, and promotion of the existing research on the relationship between intrinsic motivation and creativity. Third, Yale attitude change theory was introduced into idea implementation research, which not only expanded the application scope of this theory, but also enriched theoretical perspectives of idea implementation research.
Peer Review Status:Awaiting Review
Subjects: Psychology >> Social Psychology Subjects: Psychology >> Psychological Measurement submitted time 2019-11-15
Abstract： Social dilemma is a multi-disciplinary topic widely discussed in Ethics, Psychology, Sociology and Economics, among which moral dilemma is the typical paradigm and has a long history of research. Present article reviewed three existed empirical approaches in dilemma research, traditional dilemma paradigm, process dissociation method and Consequences-Norms-generalized Inaction/action (CNI) model. Based on analyzing the contributions and limitations of the above three approaches, we developed a Consequences-general Action/inaction preference-Norms sensitivity estimation (CAN) algorithm. With the development of the four approaches, moral dilemma research has no longer limited in the contradicted dilemma situations. CNI model and CAN algorithm have extended to consider the four combinations between proscriptive/prescriptive norms and benefits greater/smaller than costs. With CNI model and CAN algorithm, Controversies in moral theories and empirical inconsistences can be further clarified. The four approaches provided methodological references for the similar topics with potential contradictions in many other domains. Therefore, the approaches can be used in multi-disciplines."
Peer Review Status:Awaiting Review
Subjects: Psychology >> History of Psychology Subjects: Psychology >> Cognitive Psychology submitted time 2019-11-14
Abstract： Embodied cognition reveals the impacts of body and environment on cognition. Put the implications of all the embodied theories together, the relationship between effect size and effect stability of embodiment effect was different before and after mind maturity. Before mind maturity, the effect size is big but unstable; after mind maturity, the effect size is small but stable. The implications of all the theories shed light on the dual processes underlying embodiment effect. Recently, several empirical evidences also suggest the dual processes underlying embodiment effect. To summarize, continuity of being unconsciously processed, cognitive resources accessibility and potential psychological conflict are the three key boundary conditions for embodiment effect. Combining theoretical and empirical evidences, Embodied Dual Process Perspective was suggested. The basic boundary conditions and the dual process property should be addressed in light of new found embodiment effects and the replications of identified embodiment effects. The Embodied Dual Process Perspective offers a notable extension for precisely localizing the embodiment effects and advancing the therotical consideration of the replication crisis.
Peer Review Status:Awaiting Review
Subjects: Psychology >> Social Psychology submitted time 2019-09-12
Abstract： " Moral self, which is a core part of self-concept, is the overall self-evaluation in the moral domain. Previous studies in social psychology has shown that moral self is closely related to individual’s moral behaviors. Interestingly, recent studies found that moral self showed advantages over other aspects of self. For example, it was reported that self-enhancement effect in moral domain is stronger than that in other domains. However, it is still unknown whether moral self related information were processed preferentially during cognitive process. To answer this question, the current study investigated the enhancement effect of moral-self in perceptual processing, by adopting a cognitive neuroscience approach. Based on our behavioral (Chapter 2 ~ 4) and neural (Chapter 5) results, we proposed that moral self served as internal reference for individual’s cognitive processing. Firstly, the current study confirmed that the moral related information is processed preferentially during perceptual processing by three experiments (Chapter 2). Also, we excluded the cofounding factors such as familiarity of words. We reasoned that if moral self is the inner reference for information processing, then this effect should modulated by self-relatedness. Thus, if we explicitly compared the moral, neutral and immoral aspect of self and strangers, advantage of moral information should only occur on moral self. This hypothesis was confirmed by experiment 4. To further explore the interaction between morality and self-relatedness, we rendered the morality (experiment 5) or self-related (experiment 6) information as task-irrelevant information, and the interactions were found, suggesting that the moral self could be implicitly modulate the perceptual processing. To validate the stability of the effect , we conducted an meta-analysis of 6 studies, and found that the effect size of the modulation of moral self on perceptual processing is about Cohen’s d = 0.5, a moderate effect size. To further explore the mechanism underlying the facilitation effect of moral self, the current study employed cognitive modeling methods. We re-analyzed the data from the experiment 4, 5 and 6, by using drift diffusion model (DDM). The results showed that the facilitation effect of the moral-self occur primarily on drift rate, which means faster information accumulating speed. These results suggest that the perceptual salience of moral self is similar to those of physical salience. To examine the cross-task stability of the advantage of moral self, we conducted a new experiment in which participants required to finish a perceptual matching task and a perceptual categorization task. Using ex-Gaussian model and DDM, we found that the advantage of moral self is cross task stable. In the second part of this study, we explored the neural basis of moral self by Voxel-Based Morphometry (VBM) and meta-analysis. The VBM study explored the correlation between moral self-evaluation and volume of grey matters in the brain. We didn’t found any significant cluster that correlated to moral-self. We also used activation likelihood estimation (ALE) meta-analysis of fMRI studies of moral judgment and self-referential task, because these two tasks could potentially involve moral self-referential processing. Our meta-analysis found that dorsal medial frontal cortex and frontal pole are share by both moral judgment and self-referential processing, suggesting that these two brain regions may play an important role in moral self. In sum, the current study systematically investigated the cognitive mechanism and neural basis of moral self. From the perspective of computational modelling, we explore the mechanism of moral self form all three levels: function, algorithm and hardware. Our results showed that moral self is perceptual salient, and this saliency is cross-task stable. The cognitive mechanism behind this advantage lies in the higher information accumulation rate, and the dorsal medial frontal cortex and frontal pole maybe the neural basis of moral self. The study of moral self connected social psychology, cognitive psychology, social neuroscience, and computational modeling, providing solid foundation for future studies. " "
Peer Review Status:Awaiting Review
Subjects: Materials Science >> Materials Science (General) submitted time 2016-11-04 Cooperative journals: 《金属学报》
Abstract：利用三种实验方法，研究了SO2/Air/N2气氛对纯镁及AZ91D合金熔体的保护行为，运用X射线衍射仪（XRD）、扫描电镜（SEM）、能谱仪（EDS）、Auger电子能谱仪（AES）和X射线光电子能谱仪（XPS）分析了形成的表面膜的相组成、微观组织形貌和生长过程，结合热力学计算讨论了表面膜的生长与保护机理，并最终考查了表面膜的稳定性。研究表明：在SO2/Air/N2气氛中，保护性表面膜由MgO、MgS和MgSO4混合组成，MgSO4是热力学稳定相，它的形成至关重要。当采用SO2/Air/N2气氛为保护气氛且SO2含量一定时，Air含量不能过高也不能过低。
Subjects: Mathematics >> Algebra and Number Theory submitted time 2016-05-19
Abstract：For CM elliptic curve over rational field with analytic rank one, for any potential good ordinary prime p, not dividing the number of roots of unity in the complex multiplication field, we show the p-part of its Shafarevich-Tate group has order predicted by the Birch and Swinnerton-Dyer conjecture.
Peer Review Status:Awaiting Review
Hits
Hits
Downloads
Comment
YES