Deep reinforcement learning has led to dramatic breakthroughs in the field of artificial intelligence for the past few years. As the amount of rollout experience data and the size of neural networks for deep reinforce...
详细信息
Deep reinforcement learning has led to dramatic breakthroughs in the field of artificial intelligence for the past few years. As the amount of rollout experience data and the size of neural networks for deep reinforcement learning have grown continuously, handling the training process and reducing the time consumption using parallel and distributed computing is becoming an urgent and essential desire. In this article, we perform a broad and thorough investigation on training acceleration methodologies for deep reinforcement learning based on parallel and distributed computing, providing a comprehensive survey in this field with state-ofthe-art methods and pointers to core references. In particular, a taxonomy of literature is provided, along with a discussion of emerging topics and open issues. This incorporates learning system architectures, simulation parallelism, computingparallelism, distributed synchronization mechanisms, and deep evolutionary reinforcement learning. Furthermore, we compare 16 current open-source libraries and platforms with criteria of facilitating rapid development. Finally, we extrapolate future directions that deserve further research.
This paper presents a review of state-of-the-art solutions concerning the optimization of computing in the field of parallel and distributed systems. Firstly, we contribute by identifying resources and quality metrics...
详细信息
This paper presents a review of state-of-the-art solutions concerning the optimization of computing in the field of parallel and distributed systems. Firstly, we contribute by identifying resources and quality metrics in this context including servers, network interconnects, storage systems, computational devices as well as execution time/performance, energy, security, and error vulnerability, respectively. We subsequently identify commonly used problem formulations and algorithms for integer linear programming, greedy algorithms, dynamic programming, genetic algorithms, particle swarm optimization, ant colony optimization, game theory, and reinforcement learning. Afterward, we characterize frequently considered optimization problems by stating these terms in domains such as data centers, cloud, fog, blockchain, high performance, and volunteer computing. Based on the extensive analysis, we identify how particular resources and corresponding quality metrics are considered in these domains and which problem formulations are used for which system types, either parallel or distributed environments. This allows us to formulate open research problems and challenges in this field and analyze research interest in problem formulations/domains in recent years.
parallel and distributed computing (PDC) courses are useful for computer science (CS) and domain science students. For CS students, PDC is a fundamental field that examines concepts relating to a range of CS subfields...
详细信息
parallel and distributed computing (PDC) courses are useful for computer science (CS) and domain science students. For CS students, PDC is a fundamental field that examines concepts relating to a range of CS subfields, such as algorithms, architecture, simulation, software, systems, among others. Students with domain science backgrounds also require PDC to carry out their research objectives, and the ongoing data revolution has exacerbated this necessity. Given the rise of data science and other data-enabled computational fields, we propose several data-intensive pedagogic modules that are used to teach PDC using message-passing programming with the Message Passing Interface (MPI). These modules employ activities that are interesting, relevant, and accessible to both computer and domain science students enrolled in graduate level programs. Using pre-and post-module completion quizzes and anonymous free response surveys, we evaluated the efficacy of the pedagogic modules across four cohorts of students enrolled in a graduate level High Performance computing (HPC) course at Northern Arizona University. The students have diverse educational backgrounds as some students were enrolled in programs outside of CS. These programs include electrical and computer engineering, mechanical engineering, astronomy & planetary science, bioinformatics, and ecoinformatics. Despite the multidisciplinary backgrounds of the students, we find that the hands-on application-driven approach to teaching PDC was successful at helping students learn core PDC concepts, and that the modules are useful for facilitating online learning which was required during the COVID-19 pandemic.
Byzantine fault-tolerant (BFT) consensus is a critical problem in parallel and distributed computing systems, particularly with potential adversaries. Most prior work on BFT consensus assumes reliable message delivery...
详细信息
Byzantine fault-tolerant (BFT) consensus is a critical problem in parallel and distributed computing systems, particularly with potential adversaries. Most prior work on BFT consensus assumes reliable message delivery and tolerates arbitrary failures of up to n3 nodes out of n total nodes. However, many systems face unpredictable message delivery failures. This paper investigates the impact of unpredictable message delivery failures on the BFT consensus problem. We propose Nicaea, a novel protocol enabling consensus among loyal nodes when the number of Byzantine nodes is below a new threshold, given by: (2-rho)(1-rho)2n-2-1 /(2-rho)(1-rho)2n-2+1n, where rho denotes the message failure rate. Theoretical proofs and experimental results validate Nicaea's Byzantine resilience. Our findings reveal a fundamental trade-off: as message delivery instability increases, a system's tolerance to Byzantine failures decreases. The well-known n3 threshold under reliable message delivery is a special case of our generalized threshold when rho = 0. To the best of our knowledge, this work presents the first quantitative characterization of unpredictable message delivery failures' impact on Byzantine fault tolerance in parallel and distributed computing.
This paper describes two active learning strategies to teach and review fundamental PDC concepts in early computer science courses. Questions were created based on eight PDC concept categories. In the first phase, fla...
详细信息
ISBN:
(纸本)9798350311990
This paper describes two active learning strategies to teach and review fundamental PDC concepts in early computer science courses. Questions were created based on eight PDC concept categories. In the first phase, flashcards were created for students to review the concepts. In the second phase, a card game called PDC Quest was created to allow groups of students to engage collaboratively in learning and reviewing the concepts.
This special session will report on the updated NSF/IEEE-TCPP Curriculum on parallel and distributed computing released in Nov 2020 by the Center for parallel and distributed computing Curriculum Development and Educa...
详细信息
ISBN:
(纸本)9781450394338
This special session will report on the updated NSF/IEEE-TCPP Curriculum on parallel and distributed computing released in Nov 2020 by the Center for parallel and distributed computing Curriculum Development and Educational Resources (CDER). The purpose of the special session is to obtain SIGCSE community feedback on this curriculum in a highly interactive manner employing the hybrid modality and supported by a full-time CDER booth for the duration of SIGCSE. In this era of big data, cloud, and multi- and many-core systems, it is essential that the computer science (CS) and computer engineering (CE) graduates have basic skills in parallel and distributed computing (PDC). The topics are primarily organized into the areas of architecture, programming, and algorithms topics. A set of pervasive concepts that percolate across area boundaries are also identified. Version 1 of this curriculum was released in December 2012. That curriculum guideline has over 140 early adopter institutions worldwide and has been incorporated into the 2013 ACM/IEEE Computer Science curricula. This Version-II represents a major revision. The updates have focused on enhancing coverage related to the topical aspects of Big Data, Energy, and distributedcomputing. The session will also report on related CDER activities including a workshop series on a PDC institute conceptualization, developing a CE-oriented version of the curriculum, and identifying a minimal set of PDC topics aligned with ABET's exposure-level PDC requirements. The interested SIGCSE audience includes educators, authors, publishers, curriculum committee members, department chairs and administrators, professional societies, and the computing industry.
Along with high performance computer systems, the Application Programming Interface (API) used is crucial to develop efficient solutions for modern parallel and distributed computing. Compute Unified Device Architectu...
详细信息
ISBN:
(纸本)9781665440714
Along with high performance computer systems, the Application Programming Interface (API) used is crucial to develop efficient solutions for modern parallel and distributed computing. Compute Unified Device Architecture (CUDA) and Open computing Language (OpenCL) are two popular APIs that allow General Purpose Graphics Processing Unit (GPGPU, GPU for short) to accelerate processing in applications where they are supported. This paper presents a comparative study of OpenCL and CUDA and their impact on parallel and distributed computing. Mandelbrot set (represents complex numbers) generation, Marching Squares algorithm (represents embarrassingly parallelism), and Bitonic Sorting algorithm (represents distributedcomputing) are implemented using OpenCL (version 2.x) and CUDA (version 9.x) and run on a Linux-based High Performance computing (HPC) system. The HPC system uses an Intel i7-9700k processor and an Nvidia GTX 1070 GPU card. Experimental results from 25 different tests using the Mandelbrot Set generation, the Marching Squares algorithm, and the Bitonic Sorting algorithm are analyzed. According to the experimental results, CUDA performs better than OpenCL (up to 7.34x speedup). However, in most cases, OpenCL performs at an acceptable rate (CUDA speedup is less than 2x).
parallel and distributed computing (PDC) has found a broad audience that exceeds the traditional fields of computer science. This is largely due to the increasing computational demands of many engineering and domain s...
详细信息
ISBN:
(纸本)9781665435772
parallel and distributed computing (PDC) has found a broad audience that exceeds the traditional fields of computer science. This is largely due to the increasing computational demands of many engineering and domain science research objectives. Thus, there is a demonstrated need to train students with and without computer science backgrounds in core PDC concepts. Given the rise of data science and other data-enabled computational fields, we propose several data-intensive pedagogic modules that are used to teach PDC using message-passing programming with the Message Passing Interface (MPI). These modules employ activities that are common in database systems and scientific workflows that are likely to be employed by domain scientists. Our hypothesis is that using application-driven pedagogic materials facilitates student learning by providing the context needed to fully appreciate the goals of the activities. We evaluated the efficacy of using the data-intensive pedagogic modules to teach core PDC concepts using a sample of graduate students enrolled in a high performance computing course at Northern Arizona University. In the sample, only 30% of students have a traditional computer science background. We found that the hands-on application-driven approach was generally successful at helping students learn core PDC concepts.
Teaching topics related to high performance computing and parallel and distributed computing in a hands-on manner is challenging, especially at introductory, undergraduate levels. There is a participation challenge du...
详细信息
ISBN:
(纸本)9781728159751
Teaching topics related to high performance computing and parallel and distributed computing in a hands-on manner is challenging, especially at introductory, undergraduate levels. There is a participation challenge due to the need to secure access to a platform on which students can learn via hands-on activities, which is not always possible. There are also pedagogic challenges. For instance, any particular platform provided to students imposes constraints on which learning objectives can be achieved. These challenges become steeper as the topics being taught target more heterogeneous, more distributed, and/or larger platforms, as needed to prepare students for using and developing Cyberinfrastructure. To address the above challenges, we have developed a set of pedagogic activities that can be integrated piecemeal in university courses, starting at freshman levels. These activities use simulation so that students can experience hands-on any relevant application and platform scenarios. This is achieved by capitalizing on the capabilities of the WRENCH and SimGrid simulation frameworks. After describing our approach and the pedagogic activities currently available, we present results from an evaluation performed in an undergraduate university course.
Integrating parallel and distributed computing (PDC) topics in core computing courses is a topic of increasing interest for educators. However, there is a question of how best to introduce PDC to undergraduates. Sever...
详细信息
ISBN:
(纸本)9781728174457
Integrating parallel and distributed computing (PDC) topics in core computing courses is a topic of increasing interest for educators. However, there is a question of how best to introduce PDC to undergraduates. Several educators have proposed the use of "unplugged activities", such as role-playing dramatizations and analogies, to introduce PDC concepts. Yet, unplugged activities for PDC are widely-scattered and often difficult to find, making it challenging for educators to create and incorporate unplugged interventions in their classrooms. The PDCunplugged project seeks to rectify these issues by providing a free repository where educators can find and share unplugged activities related to PDC. The existing curation contains nearly forty unique unplugged activities collected from thirty years of the PDC literature and from all over the Internet, and maps each activity to relevant CS2013 PDC knowledge units and TCPP PDC topic areas. Learn more about the project at ***.
暂无评论