教学课程资源库 更多>>

关键词： Computer science  

摘要： This article describes the maps of General and secondary education, which make up a large group and display secondary schools, the provision of secondary schools with teachers, and in addition, out-of-school and children's institutions. Separately, we consider the characteristics of interactive educational maps designed to provide and solve problems of increasing the level of accessibility, visibility and convenience of presenting data about General education institutions. The algorithm for drawing up regional maps of education is described in detail. A detailed analysis of the existing maps of education (different years) in the Republic of Bashkortostan is carried out. © 2020, Journal of Computing and Information Technology. All Rights Reserved.

关键词： Computer Science   Artificial Intelligence   Data Parallelism   Model Parallelism   Hybrid Parallelism   Keras   Caffe   TensorFlow   PyTorch   MPI   Eager Execution   Deep Learning   Scalable DNN Training   MVAPICH2-GDR   CUDA-Aware MPI  

摘要： Recent advances in Machine Learning (ML) and Deep Learning (DL) techniques have triggered key success stories in many application domains like Computer Vision, Speech Comprehension and Recognition, and Natural Language Processing. Large-scale Deep Neural Networks (DNNs) are primary drivers of these success stories. However, training complex DNN architectures that consist of millions of trainable parameters is compute-intensive. Training is done using a large number of examples (training data set) and can take from weeks to months to achieve state-of-the-art prediction capabilities (accuracy). To achieve higher accuracy, making the DNN deeper and larger has become a common strategy but it also leads to a significantly bigger memory footprint. Thus, DNN training is not only compute-intensive but also a memory-hungry process requiring gigabytes of memory. To accelerate the process of large-scale DNN training, this dissertation is focused on designing high-performance systems that can exploit thousands of CPUs and GPUs for faster training. The novel approach presented in this work is called the co-design of high-performance communication middleware and DL frameworks. Co-design is necessary because of the complexity of the overall execution stack for modern DL frameworks. Broadly, this stack consists of many layers, which start from the application layer followed by the DL framework layer (e.g. TensorFlow). The next layer in the stack is the distributed training middleware layer (e.g. Horovod) that connects a DL framework to an underlying communication middleware (e.g. a Message Passing Interface (MPI) library). Finally, the communication middleware layer sits directly on top of the parallel hardware that consists of multiple CPU/GPU nodes connected with a high-performance network. \textit{The complexity of this stack coupled with inefficient existing approaches to utilize it has led to several problems}: First, there is a lack of credible and systematic performance stud

摘要： On August 31, 1970, an experiment began that continues to this day. The first chess tournament for computers was held as part of the Association for Computing Machinery's (ACM's) National Conference. The interest generated was tremendous, leading to ACM sponsoring an annual event until 1994. Chess competitions continue to this day, allowing for 50 years of data on the growth of artificial intelligence capabilities in this domain. During this period, program ratings have soared from roughly 1400 in 1970 to over 3500 today. The 1970 event was the first continuous competition in computer science history, and it represents the longest ongoing experiment in computing history.(1)

关键词： Computer science  

摘要： Explaining why an answer is in the result of a query or why it is missing from the result is important for many applications including auditing, debugging data and queries, hypothetical reasoning about data, and data exploration. Both types of questions, i.e., why and why-not provenance, have been studied extensively, but mostly in isolation. A recent study shows that unification of why and why-not provenance can be achieved by developing a provenance model for queries with negation. In many complex queries, negation is natural and yields more expressive power. Thus, supporting both types of provenance and negation together can be useful for, e.g., debugging (missing) data over complex queries with negation. However, why-not provenance and—to a lesser degree—why provenance, can be very large resulting in severe scalability and usability challenges. In this thesis, we introduce a framework that unifies why and why-not provenance. We develop a graph-based provenance model that is powerful enough to encode the evaluation of queries with negation (First-Order queries). We demonstrate that our model generalizes a wide range of provenance models from the literature. Using our model, we present the first practical approach that efficiently generates explanations, i.e., parts of the provenance that are relevant to the query outputs of interest. Furthermore, we present a novel approximate summarization technique to address the scalability and usability challenges. Our technique efficiently computes pattern-based provenance summaries that balance informativeness, conciseness, and completeness. To achieve scalability, we integrate sampling techniques into provenance capture and summarization. We implement these techniques in our PUG (Provenance Unification through Graphs) system which runs on top of a relational database. We demonstrate through extensive experiments that our approach scales to large datasets and produces comprehensive and meaningful (summaries of) provenance.

关键词： Artificial intelligence   Computer science  

摘要： A big challenge in artificial intelligence (AI) is creating autonomous agents that can interact well with other agents over extended periods of time. Most previously developed algorithms have been designed in the context of Repeated Games, environments in which the agents interact in the same scenario repeatedly. However, in most real-world interactions, relationships between people and autonomous agents consist of sequences of distinct encounters with different incentives and payoff structures. Therefore, in this thesis, we consider Interaction Games, which model interactions in which the scenario changes from encounter to encounter, often in ways that are unanticipated by the players. For example, in Interaction Games, the magnitude of payoffs as well as the structure of these payoffs can differ across encounters. Unfortunately, while there have been many algorithms developed for Repeated Games, there are no known algorithms for playing Interaction Games. Thus, we have developed two different algorithms, augmented Fictitious Play (aFP) and augmented S# (Aug-S#), for playing these games. These algorithms are designed to generalize Fictitious Play and S# algorithms, which were previously created for Repeated Games, to the more general kinds of scenarios modeled by Interaction Games. This thesis primarily focuses on the evaluation of these algorithms. We first analyze the behavioral and performance properties of these algorithms when associating with other autonomous algorithms. We then report on the results of a user study in which these algorithms were paired with people in two different Interaction Games. Our results show that while the generalized algorithms demonstrate many of the same properties in Interaction Games as they do in Repeated Games, the complexity of Interaction Games appear to alter the kinds of behaviors that are successful, particularly in environments in which communication between players is not possible.

关键词： Computer science  

摘要： In separation of source signals from signal mixtures and involving multiple sensors, an emerging useful method has been Independent Component Analysis (ICA). The other method for blind separation of source signals is Principal Component Analysis (PCA). ICA is a better method than PCA because it has no restrictions of the components required to be orthogonal to each other. Besides, ICA considers true statistical independence of the components, unlike PCA. However, in ICA, we may use PCA in one of the initial steps, so the ICA procedure is more comprehensive. ICA provides a variety of techniques to separate linearly mixed sources. We neither know what the sources are, nor do we know how they are mixed. The data is first centered and pre-whitened. One ICA technique is to estimate how they are mixed, maximizing the fourth cumulants. It leads to the estimation of non-Gaussian sources from Gaussian mixtures. Overall problem statement: Though ICA is based on strong theoretical foundations, yet it presents several implementation problems. The first important problem is the inconsistency of the components. The other problem is the difficulty in the interpretation of the groups of components in applying to tasks such as classification and prediction. This dissertation proposes to attend to these two problems. Part One: A new index to measure the inconsistency of ICA. Introduction/background: We examine one of the challenges of ICA–inconsistency or instability. The context of the study is ICA involving multiple sensors (time series data). Instability is said to occur when the independent components vary, each time when ICA is conducted. Past work: There are performance indexes as attempts to quantify the problem. State-of-the-art software packages mention of the stability issue but there is no unanimity in the choice of performance indexes; the Icasso stability index and the Amari performance index are the most frequently used. Problem statement: The problem is that the measur

关键词： Robotics   Artificial intelligence   Computer science  

摘要： Robots are becoming more of a part of our daily lives. They have become an extension of some our human capabilities and there is a need to develop control algorithms that contribute to the successful deployment of these machines to navigate, explore and map indoor human environments. These robots and their actions, despite our effort to make them as predictable as possible, show stochastic behaviors, as well as motion and sensing uncertainties. We leverage the use of Constrained Markov Decision Processes (CMDP) to balance multi-cost problems with constraints, under the premise of having multiple possible sources of uncertainties. This dissertation engages in solving some of these problems in the following chapters. Initially, we highlight some of the theoretical background about Markov Decision Process (MDP) and its extension the CMDP. From this point we deal with the problem of multiple robots visiting multiple targets, while we fix temporal and failure probability constraints. We present our solution to expand the state space following a binary sequence that represents successful observations of each of the targets. All this is classified as the rapid deployment problem, which we define and solve for a team of robots. Closing the gap between reality and theory, we implement a stochastic model that recreates the motion primitives from a robot. We proceed to use these modeled primitives to create modeled trajectories and extract transition probabilities from them. These transition probabilities characterize some of the robot’s behavior and we use them with our formulation of a CMDP. Then we calculate a navigation policy to traverse some real scenarios. We create and implement a new spatial model dubbed Oriented Topological Semantic Map (OTSM). This new type of map can be built in run-time, and together with a CMDP, we assign actionable temporal deadlines to the robot executing an exploration task. We open-sourced a ROS framework that can be downloaded and used to re

关键词： Artificial intelligence   Computer science  

摘要： Smart cities are envisioned to include million of sensors and devices tied together through the Internet. The sensors generate huge amount of data that can be potentially used to develop several applications. Connected vehicles can be considered as a significant realm of the smart city. Furthermore, among connected vehicles, Plug-in Electric Vehicles (PEVs) are becoming an integral component of smart city. Plug-in Electric Vehicles (PEVs) can play a major role in reducing carbon footprints in transportation sector. However, with the growing increase in the usage of PEVs, their charging infrastructure need to be comprehensively developed. This thesis develops optimization strategies for parked and en route PEV charging. First, the thesis develops a charging strategy for parked vehicles. For parked vehicle charging, electricity demand must have less variation at different times in a day. This helps in avoiding any detrimental effect on power electronic equipment, reduces electricity cost, and simplifies the electric energy demand prediction from the smart grid. This thesis delineates an intelligent aggregator architecture to automate parked Plug-in Electric Vehicle (PEV) charging in large parking places to reduce dynamic load variation. The thesis presents a set of novel rectangles placement-based algorithms to schedule the PEV charging based on their arrival, departure time at parking place, and their charging requirement. En route PEVs have different requirements as compared to parked PEVs. To facilitate en route PEV charging, this research investigates an integrated COP architecture. The COP constitutes a Communication unit that provides network management solution for PEV charging, an Optimization unit for allocating charging station to PEVs, and a Prediction unit to determine traffic flow information in advance. The thesis shows how three units can interact with each other to provide efficient charging infrastructure to PEVs. A related problem regarding applicati

关键词： Computer science  

摘要： Pathfinding is the search for an optimal path from a start location to a goal location in a given environment. In Artificial Intelligence pathfinding algorithms are typically designed as a kind of graph search. These algorithms are applicable in a wide variety of applications such as computer games, robotics, networks, and navigation systems. The performance of these algorithms is affected by several factors such as the problem size, path length, the number and distribution of obstacles, data structures and heuristics. When new pathfinding algorithms are proposed in the literature, their performance is often investigated empirically (if at all). Proper experimental design and analysis is crucial to provide an informative and non- misleading evaluation. In this research, we survey many papers and classify them according to their methodology, experimental design, and analytical techniques. We identify some weaknesses in these areas that are all too frequently found in reported approaches. We first found the pitfalls in pathfinding research and then provide solutions by creating example problems. Our research shows that spurious effects, control conditions provide solutions to avoid these pitfalls.

关键词： Information technology   Computer science  

摘要： Cyber attacks are on the rise, with potentially devastating effects at the personal, business, and national levels. Despite real and evolving cyber threats, people often fail to implement and effectively use basic, well-known cybersecurity technologies and practices. Further contributing to the cybersecurity problem is the shortage of security personnel to address these challenges. A critical role and force-multiplier in security adoption is the cybersecurity advocate: a security professional who has the skills to effectively promote security and facilitate positive security behavior change. Cybersecurity advocates attempt to remedy implementation failures by promoting and facilitating the adoption of security best practices and technologies as an integral component of their jobs. Currently, there is no clear career track and few resources for educating professionals on how to be good cybersecurity advocates. Furthermore, it is unclear as to what advocacy techniques may be most effective. In addition to the bias towards technical skills, these gaps are likely due to the fact that we have little understanding of the work practices and competencies that lead to successful security advocacy. The purpose of my research is to gain a better understanding of these work practices. A first stage in my investigation involved interviews of professional security advocates. Since this interview data was one-sided from the perspective of advocates themselves, I validated the findings with a second stage exploring the effectiveness of advocates’ approaches via a case study of a security awareness team at a U.S. government agency. This research uncovers definitional boundaries of cybersecurity advocates, including skills, characteristics, motivations, challenges, and tactics. Findings reveal that advocates employ technical and non-technical skills and a variety of techniques to overcome negative perceptions of security and other barriers to security adoption. A better understanding