教学课程资源库 更多>>

关键词： Computer science   Artificial intelligence  

摘要： The speed with which new online content becomes available has exacerbated the well known problem of information overload and motivated the innovation of techniques that help people read and consume information. More specifically for NLP, it has motivated researchers to design general models that can extract from large collections of documents informative structured information, such as information about events – what’s happening around the world. The extracted structured information (e.g., participants, locations, objects involved in an event) is essential for a variety of downstream tasks such as knowledge base population, question answering, and document analysis. In recent years, with the progress in research on deep learning, the community has seen improvements on many sentence-level information extraction tasks such as named entity recognition and relation extraction. But less progress has been made on document-level extraction problems (where the elements to be extracted are spread across the document), despite the fact that document-level extraction is closer to what is needed by end users. Existing methods largely ignore the document-level context and split the full extraction problem into separate tasks which cause error propagation. Plus, they rely heavily on manually annotated resources developed for a fixed domain-specific output schema, and, as a result, are not data-efficient or general enough to handle unanticipated schema changes at deployment time. In this dissertation, we introduce models and frameworks to address these shortcomings of prior work. To better incorporate the document-level context we propose a multi-granularity machine reader, which interprets sentences in the context of preceding and following sentences. To help neural network-based models better capture the output structure and dependencies between events, we propose a generative learning-based framework for the extraction problem, which tackles this complicated task in one pass, avoidi

关键词： Industrial engineering   Operations research   Computer science  

摘要： In recent decades, there has been significant growth in the development of rechargeable battery-powered devices such as electric vehicles, leading to an urgent demand for batteries with high reliability and quality. The end-of-life (EoL), a critical indicator of battery health, can be estimated by adaptive stochastic processes or advanced machine learning techniques. However, unless such approaches assume that the degradation path has a specific form, they operate as black boxes and are unable to provide stochastic interpretation. To address these challenges, a pattern-driven degradation process (PdDP) was executed that can model battery degradation, controlling degradation fluctuation by a GRU-driven degradation pattern. Further, a joint-learning sampling-based expectation maximization (JSEM) algorithm was developed to handle non-Markovian state transitions in estimating model parameters. Finally, a case study showed that the proposed methods outperform traditional methods with respect to both one-step and multistep-ahead prediction accuracies.

关键词： Algorithms   Linear algebra   Computer science   Mathematics  

摘要： The Kalman filter and Kalman smoother are important components in modern multitarget tracking systems. Their application are vast which include guidance, navigation and control of vehicles. On top of that, when the motion model is uncertain, MultipleModel approach can be combined with the filtering and smoothing method. However, with large amount of retrodiction window size, number of motion models and large number of targets, this process can become very computationally intensive and thus time consuming. Very often, real-time processing is needed in the world of tracking and therefore, this computational bottleneck become a problem. This is the motivation behind this thesis, to reduce the computational complexity when multi-target, multiwindow or multi-model applications are used. This thesis presents several approaches to tackle this multi-dimensional problems in terms of complexity while maintaining satisfactory precision. A natural step forward will be in leveraging the modern multi-core architectures. However, in order to parallelize such process, these algorithms have to be reformulated to be fitted into the parallel processors. In order to parallelise multi-target and multi-window scenario, this thesis introduce nested parallelism and prefix-sum algorithm to tackle the problem and realised this on Intel Knights Landing (KNL) Processor and OpenMP memory model. On the other hand, in the case of limited parallel resources, this thesis also develop alternatives called Fast Kalman smoother (FRTS) to lower the computation complexity due to multi-window problem. Specifically the smoother algorithm is reformulated such that it is computationally independent of number of window size in the fixed-lag configuration. Although the underlying mathematics is the same as the conventional approach, FRTS introduced numerical stability issue which makes the smoother unstable. Therefore, this thesis introduce the idea of condition number to monitor the deterioration rate in orde

关键词： Computer science   Artificial intelligence  

关键词： Image retrieval   Training   Semantics   Hamming distance   Computer science   Laplace equations   Image Retrieval   Hashing   Non-stationary Environment   Concept Drift   Re-ranking  

摘要： Concept drift is prevalent in non-stationary data environments but is rarely researched in image retrieval. Therefore, more research is needed on image retrieval in non-stationary data environments so that highly relevant images can still be retrieved when concept drifts happen. Hashing is a key technique to allow efficient image retrieval, so incremental hashing technique emerges in recent years for image retrieval in non-stationary environments. A state-of-the-art method is Incremental Hashing (ICH). ICH trains new hash tables on new data without considering the performance of previous hash tables, so the dependency of successive hash tables is ignored. To make use of this dependency in order to improve the performance of image retrieval in non-stationary environments, Complementary Incremental Hashing with query-adaptive Re-ranking (CIHR) is proposed in this paper. CIHR trains multiple hash tables incrementally, one for each data chunk of images. A new hash table is trained on a new data chunk of images as well as those images badly hashed by previous hash tables, thus the new hash table is complementary to the previous hash tables. To use the hash tables more effectively, a query-adaptive re-ranking method is used to weight all hash functions in each hash table according to their retrieval performance with respect to a given query. Weighted Hamming distance is finally used to evaluate the similarity between the query and the images in the database, as the basis of image retrieval. Experimental results on simulated non-stationary scenarios show that the proposed CIHR method achieves higher retrieval accuracy than all methods being compared, thus setting a new state of the art in image retrieval in non-stationary data environments.

关键词： Digital Learning   Middle East   TPACK Model   Web-Enhanced Course  

摘要： Information and communication technologies have changed the way students learn, and higher education institutions in the Middle East region strive to attract professors who use technology in their daily teaching and learning practices. This study examines students' experiences in a webenhanced computer science course. In an online survey based on the technological pedagogical content knowledge (TPACK) model, 132 students' responses to 31 items were measured and analyzed. Factor analysis revealed three pre-dominated factors: technology, pedagogy-content, and TPACK;of these, technology was found to exert the strongest influence on students' educational experiences. The findings are discussed in relation to the TPACK model, along with implications for theory and practice. This investigation provides insights into current initiatives in the Middle East region that encourage higher education institutions to implement technology in teaching and learning. It contributes to the literature by deepening understanding of how new approaches in education are implemented and adopted.

关键词： Application programming interface   Personal computers   Performance evaluation   Artificial intelligence   Magnetic resonance imaging   Neural networks   Field programmable gate arrays   Computer science   Medical imaging  

摘要： Görüntü işleme; savunma sanayisi, biyomedikal, tüketici elektroniği gibi sektörlerde uygulamaları olan bir araştırma alanıdır. Zamanla, kullanılan algoritmaların karmaşıklığı ve kullanılan görüntülerin çözünürlüğü artmıştır. Buna karşın, karmaşıklığı fazla algoritmaların çalıştırılmasının ya da büyük boyutlu görüntülerin işlenmesinin tek çekirdekli merkezi işlem birimleri ile yeteri kadar hızlı bir şekilde yapılamadığı görülmüştür. Bunun sonucunda çok çekirdekli merkezi işlem birimleri kullanılmaya başlanmıştır fakat bunun da yetersiz kaldığının fark edilmesi üzerine yüzlerce aritmetik ve mantık işlem birimi barından grafik işleme birimlerinin görüntü işleme alanında kullanımı önem kazanmıştır. Grafik işleme birimleri de standart mikrodenetleyiciler gibi alt-seviyeli ve üst-seviyeli bazı programlama ara yüzleri ile programlanabilmektedir. Alt-seviyeli programlama ara yüzleri yüksek performans getirileri sunsa da bu programların geliştirilmesi zamanın yanında yüksek miktarda donanım bilgisi de gerektirmektedir bu da araştırmacılara zorluk çıkarmaktadır. Bu tez çalışmasında amaç, altseviyeli bir ara yüz kullanmadan, sadece üst-seviyeli bir ara yüz kullanarak görüntü işleme operasyonlarını grafik işleme birimi üzerinde çalıştırmak ve üst-seviyeli ara yüzlerin de kayda değer bir hızlanma sağladığını göstermek, böylece araştırmacıları çalışmalarında üst-seviyeli ara yüzlerle grafik işlemcilerini programlamaya teşvik etmektir. Çalışma dahilinde literatürden seçilen bir görüntü işleme yöntemi, OpenMP uygulama programlama ara yüzü ve CUDA temelli OpenCV uygulama programlama ara yüzü ile C++ programlama dili kullanılarak uygulanmıştır. Uygulama öncelikle kırk sekiz çekirdekli bir bulut bilgisayarında çalıştırılarak merkezi işlem biriminin performansı ölçülmüştür. Aynı uygulama daha sonra NVIDIA GeForce 1050 GTX TI grafik kartı ve Intel i7-7700HQ merkezi işlem birimi barındıran bir kişisel bilgisayar üzerinde çalıştırılmış ve performans sonuçları karşılaştırılmıştır. Sonuçlar

关键词： Artificial intelligence   Computer science   Mathematics   Materials science   Mechanical engineering  

摘要： Zeroth-Order (ZO) methods have been applied to solve black-box or simulation-based optimization prroblems. These problems arise in many important applications nowa- days, e.g., generating adversarial attacks on machine learning systems, learning to control the system with complicated physics structure or human in the loop. In these problem settings, the objective function to optimize does not have an explicit math- ematical form and therefore its gradient cannot be obtained. This invalidates all gradient-based optimization approaches. On the other hand, ZO methods approxi- mate the gradient by using the objective function values. Many existing ZO methods adopt the two-point feedback scheme to approximate the unknown gradient due to its low estimation variance and fast convergence speed. Specifically, two-point ZO method estimates the gradient at the current iterate of the algorithm by querying the objective function value twice at two distinct neighbor points around the current iterate. Such scheme becomes infeasible or difficult to implement when the objective function is time-varying, or when multiple agents collaboratively optimize a global objective function that depends on all agents’ decisions, because the value of the objective function can be queried only once at a single decision point. However, con- ventional ZO method based on one-point feedback is subject to large variance of the gradient estimation and therefore slows down the convergence. In this dissertation, we propose a novel one-point ZO method based on the residual feedback. Specifically, the residual feedback scheme estimates the gradient using the residual between the values of the objective function at two consecutive iterates of the algorithm. When optimizing a deterministic Lipschitz function, we show that the query complexity of ZO with the proposed one-point residual feedback matches that of ZO with the existing two-point schemes. Moreover, the query complexity of the proposed algorithm can

关键词： Analytical chemistry   Atomic physics   Biology   Chemistry   Computer science   Energy   Geography   Mechanics   Physics  

摘要： Electrical resistivity is one characteristic and important physical property of a metal, and it is sensitive to the composition and microstructure. The relationship between resistivity and composition and microstructure makes resistivity a useful tool in materials research, such as non-destructive evaluation and monitoring precipitation kinetics. However, this needs a good understanding of how composition and microstructure influence resistivity, which is currently lack in Mg alloys. Therefore, a systematic investigation of the resistivity of Mg alloys is necessary. Mg-Al, Mg-Gd, Mg-Sn and Mg-Zn series alloys with different solute content are prepared for the current investigation. The resistivity of these alloys in the as-cast, solution treated, and aged status are measured at different temperatures to study the influence of temperature, composition and microstructure on the resistivity. In situ measurements are also conducted to study the resistivity changes during isothermal ageing of Mg alloys. The results show that Mg alloys have a positive temperature coefficient of resistivity (TCR). The TCR varies from different solute content, which demonstrate the deviation from Matthiessen’s rule in Mg alloys. When the alloys are solution treated, the following equation can describe the relationship between resistivity and solute contents: ρ(T)=ρMg(T)+δ(T)×c ρ(T)is the resistivity of the alloy under a certain temperature, ρMg(T) is the resistivity of pure Mg, δ(T) is the coefficient, and c is the concentration of the solute. δ(T) depends on both the temperature and the type of solute. The reason for the increment is the lattice distortion caused by the solute elements. When the alloys are aged, a phenomenological formula can describe the relationship between the resistivity and the volume fraction of precipitates: ρeff=ρα1+1/2Vβ/1-Vβ ρeffis the effective resistivity, ???? is the volume fraction of the precipitates, ρα is the resistivity of the α-Mg matrix. With the help o

关键词： Neurons   Algorithms   Twins   Neural networks   Artificial intelligence   Computer science  

摘要： There is great interest in the machine learning community to reduce Deep Neural Network (DNN) model sizes. Decreasing the memory and compute requirements expands the range of resource-constrained mobile applications where DNNs can be deployed. By far, the most popular method of compressing model size is uniform quantization. In this work, we illustrate how quantization performance was fortuitously advanced by Batch Normalization (BatchNorm), a technique originally developed to aid training convergence. This improvement is due to BatchNorm's reshaping of the network's activation distributions. Additionally, due to the limited consensus on why BatchNorm is effective, this work uses concepts from the traditional adaptive filter domain to provide insights into its dynamics and inner workings. First, we show that the convolution weight updates have natural modes whose stability and convergence speed are tied to the eigenvalues of the input autocorrelation matrices. Furthermore, our experiments demonstrate that the speed and stability benefits are distinct effects. At low learning rates, it is BatchNorm's amplification of the smallest eigenvalues that improves convergence speed. In contrast, at high learning rates, it is BatchNorm's suppression of the largest eigenvalues that ensures stability. Next, we prove that in the first training step, when normalization is needed most, BatchNorm satisfies the same optimization as Normalized Least Mean Square (NLMS), while it continues to approximate this condition in subsequent steps. The analyses provided lay the groundwork for gaining further insight into the operation of modern neural network structures using adaptive filter theory. Finally, we highlight contributions made to a real-world application of DNNs in the Smart Hospital space.