教学课程资源库 更多>>

关键词： Teaching   Construction   Software   Geometry   Civil engineering   Computer science   Education   Engineering   Mathematics  

关键词： Robotics   Artificial intelligence   Computer science  

摘要： Indoor localization through natural language has realistic significance. It is also a difficult area for human-robot interaction and collaboration. This project aims to address these issues. First, this project generates datasets that can be used for indoor localization task based on existing datasets applied to indoor navigation task. Then, various models based on Bayes rule and neural networks are proposed and their performances are verified. Through comparison, it is found that the Bayesian-based model is faster to train and process. However, its generalization capability is constrained by its property of requiring advance indexing of each viewpoint. In contrast, neural networks-based models have an advantage in generalization. By introducing attention mechanism, the stacked cross attention network (SCAN) performs better on large dataset. This project provides directions and guidance for further research in this area.

关键词： Gaussian noise   Cryptography   Matrices   Encryption   Protocols   Licenses   Computer science   Learning with errors   learning with rounding   multi-key FHE  

摘要： Fully dynamic multi-key fully homomorphic encryption (FHE) that allows an unlimited number of homomorphic operations for unconstrained parties. That is to say, it supports performing as many computational procedures on inputs (which are encrypted by an unrestricted number of parties) as needed. The existed fully dynamic multi-key FHE scheme is based on the learning with errors (LWE) problem. However, the LWE problem suffers from the intricate and laborious Gaussian sampling which extremely weakens the schemes' efficiencies. What's more, it has been lately revealed that Gaussian sampling creates lots of possible side-channel vulnerabilities that result in the disclosure of secret keys. The paper proposed a fully dynamic multi-key FHE that based on the learning with rounding (LWR) problem which leaves out the time-consuming Gaussian sampling required in the LWE problem while sustaining almost the same security level.

关键词： Computer science   Artificial intelligence   Software   Dictionaries   Datasets   Memory   Biometrics   Video recordings   Sensors   Noise   Methods   Algorithms   Latent fingerprints   Feedback   Critics   Security systems   Skills  

摘要： In 2018, a paper by P. Bontrager et al. demonstrated that individual synthetic fingerprint images generated by generative adversarial networks (GANs) are capable of matching a given fingerprint as much as 30 times more than a random single human fingerprint. In addition to examining the potential implications and usages of such artificial fingerprints, this thesis discusses some of the types of attacks possible on fingerprint-based authentication systems. These discussions serve as a backdrop for the primary focus, which is a description of the processes behind the development of implementations of systems based on the system outlined in the paper by P. Bontrager et al. to facilitate an understanding of different types of GANs, the characteristics and results of these systems, and their combined contribution to the implementation of an evolved form of GAN trained on a portion of the fingerprint images contained within the National Institute of Standards and Technology's Special Database 302.

关键词： Mathematics   Computer science  

摘要： In this work we study the effects of cheap-talk in games with rational agents. We begin by showing that all asynchronous interactions between n players and a mediator can be t-bisimulated without the mediator if n > 4t. Intuitively, t-bisimulation means that for any deviation performed by an adversary controlling at most t players in the scenario with the mediator or in the scenario without the mediator there exists an equivalent deviation in the other scenario (i.e., a deviation that produces the same outcome). We then use this result to show that any (k,t)-robust strategy in a game with n players and a mediator in an asynchronous system can be implemented with cheap talk if n > 4k + 4t. A (k,t)-robust strategy is one in which no coalition of t malicious players can decrease the payoff of anyone else and no coalition of k players can increase their payoff even in coalition with the t malicious players. We also prove that the result above can be satisfied if n > 3k + 4t whenever honest players can punish players that are caught deviating. A similar result can be shown for synchronous systems, but in this case we only require that n > 2k + 3t. We also show that for every protocol →π for n players and all k < n there exists a belief system that is consistent with →π in which all coalitions k of at most k players believe that, if there is any deviation during the execution of the protocol, then it was someone sending an incorrect message to a player in k. We call these beliefs k-paranoid. Intuitively, k-paranoid beliefs are such that all coalitions of size at most k believe that the remaining players are being honest between them. We use these beliefs to extend the results regarding (k,t)-robustness by showing that all k-resilient sequential equilibria with a mediator can be implemented with cheap-talk if n > 4k in asynchronous systems or n > 3k in synchronous systems. We finish this work by proving a matching lower bound for most of our results.

关键词： Photovoltaic cells   Solar energy   Algorithms   Alternative energy sources   Energy resources   Renewable resources   Sensors   Alternative energy   Cellular biology   Computer science   Energy  

摘要： When a Photovoltaic (PV) system is partially shaded in the environment, the current-voltage (I-V) and power-voltage (P-V) curves exhibit multiple stairs/peaks and the locus of Maximum Power Point (MPP) varies over a wide range. Such Partial Shading Conditions (PSC) bring challenges to the Maximum Power Point Tracking (MPPT) systems. This thesis presents some novel shading information to characterize the complex PSC and MPPT techniques based on the shading perception. Shading information is the mathematical indicator to express the shading patterns. The existing shading information, such as shading rate and shading strength, has the limitations that they can only characterize the PSC with two irradiation levels. To improve the application range of the shading information, the shading matrix and shading vector are proposed in this thesis. The identification and detection methods for the proposed shading information are also included. Results from simulations and experiments have shown the effectiveness and accuracy of the proposed shading detection methods. Under PSC, the power characteristics of the PV systems are too complicated that there exist multiple MPPs. The traditional MPPT techniques may be trapped in the Local MPPs (LMPPs) instead of the Global MPP (GMPP). In this thesis, some novel methods are proposed to estimate the GMPP location from the detected shading information. The proposed MPPT techniques based on the shading perception are capable of tracking the GMPP fast and accurately. Simulations and experiments are conducted to validate the performance of the proposed MPPT methods with the comparison with some well-known MPPT methods.

关键词： Computer science   Biomedical engineering   Artificial intelligence   Aging   Internet of Things   Fourier transforms   Radar   Intelligence   Neural networks  

摘要： Dans les dernières années, la population mondiale a souffert d’un vieillissement. Cette tendance vise à croître dans les prochaines années. Un des problèmes qui en découle est que le nombre de personnes nécessitant du soutien au quotidien devrait également croître avec le vieillissement de la population. Les habitats intelligents offrent une alternative intéressante pouvant soulager une partie de ce problème. Les habitats intelligents contiennent une panoplie de capteurs qui, combinés à l’intelligence artificielle, ont le potentiel de supporter leurs occupants dans leur quotidien. Au centre de cette solution technologique se trouve la reconnaissance d’activités. La reconnaissance d’activités est un problème complexe visant à identifier quelle activité est effectuée par une personne. La complexité de ce problème découle de la nature des activités à identifier, des capteurs disponibles et de l’impact de ces derniers sur la vie privée. Parmi les capteurs exploitables pour la reconnaissance d’activités, les radars Ultra-Wideband (UWB) ont généré beaucoup d’intérêt dans les dernières années, car ils offrent un compromis intéressant entre précision et protection de la vie privée. Les travaux couverts par ce mémoire visent à atteindre un meilleur taux de reconnaissance d’activités de la vie quotidienne que les travaux précédents en combinant réseaux de neurones et radar UWB. Pour ce faire, deux approches ont été testées. Dans un premier temps, la combinaison de système experts et de réseau de neurones est explorée. Ensuite, un nouveau modèle d’apprentissage profond combinant différentes architectures (Long Short-Term Memory (LSTM) et réseau de neurones convolutif (CNN)) est développé pour faire la reconnaissance d’activités. Ce modèle a comme avantage de fusionner des données provenant de différentes sources et exploite des données prétraitées minimalement. Ce modèle, appelé Tuned EfficientNetB0 avec LSTM, offre une amélioration de 18,63 % du taux de reconnaissance par rap

关键词： Computer science   Computer engineering   Artificial intelligence  

摘要： In the world of medical informatics, biomarkers play a pivotal role in determining the physical state of human beings, distinguishing the pathologic from the clinically normal. In recent years, behavioral markers, due to their availability and low cost, have attracted a lot of attention as a potential supplement to biomarkers. “Language markers” such as spoken words and lexical preference have been shown to be both cost-effective as well as predictive of complex diseases such as mild cognitive impairment (MCI). However, language markers, although universal, do not possess many of the favorable properties that characterize traditional biomakers. For example, different people may exhibit similar use of language under certain conversational contexts (non-unique), and a person's lexical preferences may change over time (non-stationary). As a result, it is unclear whether any set of language markers can be measured in a consistent manner. My thesis projects provide solutions to some of the limitations of language markers: (1) We formalize the problem of learning a dialog policy to measure language markers as an optimization problem which we call persona authentication. We provide a learning algorithm for finding such a dialog policy that can generalize to unseen personalities. (2) We apply our dialog policy framework on real-world data for MCI prediction and show that the proposed pipeline improves prediction against supervised learning baselines. (3) To address non-stationarity, we introduce an effective way to do temporally-dependent and non-i.i.d. feature selection through an adversarial learning framework which we call precision sensing. (4) Finally, on the prediction side, we propose a method for improving the sample efficiency of classifiers by retaining privileged information (auxiliary features available only at training time).

关键词： Offshore drilling   Friction   Soils   Finite element analysis   Algorithms   Density   Geometry   Shear strength   Computer science   Mathematics   Soil sciences   Environmental engineering   Civil engineering  

摘要： The installation of mobile jack-up spudcan foundations may induce severe stresses on the piles supporting the adjacent permanent jacket platforms. Existing codes (SNAME, ISO, ABS) provide scanty guidelines in assessing spudcan-pile interaction in single clay with no recommendations for layered soil profiles. In this thesis, both simplified numerical solutions and large deformation finite element (LDFE) analyses are employed to assess spudcan-pile interaction in single soft clay, clay overlying sand, and sand overlying clay profiles. These cases entail the commonly encountered jack-up installation problems including deep spudcan penetration depth, soil squeezing, and potential punch-through failure, and their impact on pile foundations in close proximity. This thesis proposes an efficient cost-effective numerical model combining analytical and numerical modeling to assess spudcan-pile interaction in single soft clay, clay overlying sand, and sand overlying clay. Using both the proposed simplified numerical model and LDFE analyses developed in this thesis, the investigation of the interaction reveals the fundamental mechanisms of spudcan-pile interactions in single and layered soils.

关键词： Mathematics   Applied mathematics   Computer science