In wireless sensor networks (WSN), reliable monitoring of a phenomenon (or event detection) depends on the collective data provided by the target cluster of sensors and not on any individual node. In this paper we def...
详细信息
A system for performing highly parallel, predictable real-time image filtration is described. The application software was developed using a special class of abstract data types (ADT) constructed in the RT-X language ...
详细信息
Although platform-independent runtime systems for parallel programming languages are desirable, the need for low-level optimizations usually precludes their existence. This is because most optimizations involve some c...
详细信息
Although platform-independent runtime systems for parallel programming languages are desirable, the need for low-level optimizations usually precludes their existence. This is because most optimizations involve some combination of low-level communication and low-level threading the product of which is almost always platform-dependent. We propose a solution to the threading half of this dilemma by using a thread package, that allows fine-grain control over the behaviour of the threads while still providing performance comparable to hand-tuned, machine-dependent thread packages. This makes it possible to construct platform-independent thread modules for parallel runtime systems and, more importantly, to optimize them.
Due to its simplicity and scalability, the differentiated services (DiffServ) model is expected to be widely deployed across the Internet. For each DiffServ compliant router, the scheduling algorithm is critical in im...
详细信息
Due to its simplicity and scalability, the differentiated services (DiffServ) model is expected to be widely deployed across the Internet. For each DiffServ compliant router, the scheduling algorithm is critical in implementing per hop behaviors (PHBs), according to which packets are forwarded. In this paper, we propose the hierarchical DiffServ scheduling (HDS) algorithm to support DiffServ classes on input-queued switches. The proposed HDS algorithm features in a hierarchical scheduling scheme that consists of two levels of schedulers. One level is the central scheduler which is designed to maximize the switch throughput by computing a maximal size matching between input ports and output ports. The other level is formed by input port schedulers which provide differentiated services by serving cells belonging to different classes dynamically. Using such a hierarchical scheme, the implementation complexity and the amount of information needs to be transmitted between input ports and the central scheduler are dramatically reduced compared with existing maximal weight matching based DiffServ scheduling algorithms. The tradeoff of its slightly worse delay performance is acceptable.
The paper aims to obtain existence and uniqueness of the solution as well as asymptotic estimate of the solution for singularly perturbed nonlinear thirdorder Robin boundary value problem with a turning point. In orde...
详细信息
The paper aims to obtain existence and uniqueness of the solution as well as asymptotic estimate of the solution for singularly perturbed nonlinear thirdorder Robin boundary value problem with a turning point. In order to achieve this aim, existence and uniqueness of the solution for third-order nonlinear Robin boundary value problem is derived first based on the upper and lower solutions method under relatively weaker conditions. In this manner, the goal of this paper is gained by applying the existence and uniqueness results mentioned above.
Informally, a grassroots system is a distributed system that can have multiple instances, independent of each other and of any global resources, that can interoperate once interconnected. More formally, in a grassroot...
详细信息
Informally, a grassroots system is a distributed system that can have multiple instances, independent of each other and of any global resources, that can interoperate once interconnected. More formally, in a grassroots system the set of all correct behaviors of a set of agents P is strictly included in the set of the correct behaviors of P when embedded within a larger set of agents P′ ⊃ P: Included, meaning that members of P are correct to ignore outside agents indefinitely;strictly included, meaning that the interaction of members of P with outside agents is also a possible and correct behavior. Client-server/cloud computing systems are not grassroots, and neither are systems designed to have a single global instance (Bitcoin/Ethereum with hardwired seed miners/bootnodes), and systems that rely on a single global data structure (IPFS, DHTs). An example grassroots system would be a serverless smartphone-based social network supporting multiple independently-budding communities that can merge when a member of one community becomes also a member of another. Grassroots applications are potentially important as they may allow people to conduct their social, economic, civic, and political lives in the digital realm solely using the networked computing devices they own and operate (e.g., smartphones), free of third-party control, surveillance, manipulation, coercion, or rent seeking (e.g., by global digital platforms such as Facebook or Bitcoin). Here, we formalize the notion of grassroots systems and grassroots implementations;specify an abstract grassroots dissemination protocol;describe and prove an implementation of grassroots dissemination for the model of asynchrony;extend the implementation to mobile (address-changing) devices that communicate via an unreliable network (e.g. smartphones using UDP);and discuss how grassroots dissemination can realize applications that support digital sovereignty – grassroots social networking and grassroots currencies. The math.mat
作者:
Aleksandr BeznosikovSamuel HorváthPeter RichtárikMher SafaryanComputer
Electrical and Math. Sciences and Engineering Division King Abdullah University of Science and Technology Thuwal KSA and Skolkovo Institute of Science and Technology Moscow Russia and School of Applied Mathematics and Informatics Moscow Institute of Physics and Technology Moscow Russia Computer
Electrical and Math. Sciences and Engineering Division King Abdullah University of Science and Technology Thuwal KSA
In the last few years, various communication compression techniques have emerged as an indispensable tool helping to alleviate the communication bottleneck in distributed learning. However, despite the fact biased com...
详细信息
In the last few years, various communication compression techniques have emerged as an indispensable tool helping to alleviate the communication bottleneck in distributed learning. However, despite the fact biased compressors often show superior performance in practice when compared to the much more studied and understood unbiased compressors, very little is known about them. In this work we study three classes of biased compression operators, two of which are new, and their performance when applied to (stochastic) gradient descent and distributed (stochastic) gradient descent. We show for the first time that biased compressors can lead to linear convergence rates both in the single node and distributed settings. We prove that distributed compressed SGD method, employed with error feedback mechanism, enjoys the ergodic rate $O\left( \delta L \exp[-\frac{\mu K}{\delta L}] + \frac{(C + \delta D)}{K\mu}\right)$, where δ ≥1 is a compression parameter which grows when more compression is applied, L and µ are the smoothness and strong convexity constants, C captures stochastic gradient noise (C = 0 if full gradients are computed on each node) and D captures the variance of the gradients at the optimum (D = 0 for over-parameterized models). Further, via a theoretical study of several synthetic and empirical distributions of communicated gradients, we shed light on why and by how much biased compressors outperform their unbiased variants. Finally, we propose several new biased compressors with promising theoretical guarantees and practical performance.
Deals with the problem of enhancing the versatility of VLSI processor arrays without undue addition of hardware, time/control overhead, and software complexity. A promising approach to this problem is based on matrix ...
详细信息
Deals with the problem of enhancing the versatility of VLSI processor arrays without undue addition of hardware, time/control overhead, and software complexity. A promising approach to this problem is based on matrix computations carried out through the Faddeev algorithm. We design a fixed-size, linear array architecture with fully local communications and straightforward control requirements. This high-throughput, systolic-type architecture allows us to minimize both I/O requirements and the number of processing elements performing complicated operations like divisions. To derive the array from a formal description of the Faddeev algorithm based on Gaussian elimination with partial pivoting, we use purposive transformations of the basic dependence graph of the algorithm before its space-time mappings onto array architectures.< >
A method is proposed to generate homoclinic solutions for an integrable nonlinear PDE with periodic boundaries. This approach resembles the dressing method known in the theory of solitons. The pole positions in the dr...
详细信息
Learning in the presence of data imbalances presents a great challenge to machine learning. Imbalanced data sets represent a significant problem because the corresponding classifier has a tendency to ignore samples wh...
详细信息
暂无评论