The Experts below are selected from a list of 159 Experts worldwide ranked by ideXlab platform

Wiroonsak Santipach - One of the best experts on this subject based on the ideXlab platform.

  • Signature Quantization in Fading CDMA with Limited Feedback
    IEEE Transactions on Communications, 2011
    Co-Authors: Wiroonsak Santipach
    Abstract:

    In this work, we analyze the performance of a signature quantization scheme for reverse-link Direct Sequence (DS)- Code Division Multiple Access (CDMA). Assuming perfect estimates of the channel and Interference covariance, the receiver selects the signature that Minimizes Interference power or maximizes signal-to-Interference plus noise ratio (SINR) for a desired user from a signature codebook. The codebook index corresponding to the optimal signature is then relayed to the user with a finite number of bits via a feedback channel. Here we are interested in the performance of a Random Vector Quantization (RVQ) codebook, which contains independent isotropically distributed vectors. Assuming arbitrary transmit power allocation, we consider additive white Gaussian noise (AWGN) channel first with no fading and subsequently, with multipath fading. We derive the corresponding SINR in a large system limit at the output of matched filter and linear minimum mean squared error (MMSE) receiver. Numerical examples show that the derived large system results give a good approximation to the performance of finite-size system and that the MMSE receiver achieves close to a single-user performance with only one feedback bit per signature element.

  • Kd-tree codebook for limited feedback CDMA
    2010 17th International Conference on Telecommunications, 2010
    Co-Authors: Kritsada Mamat, Wiroonsak Santipach
    Abstract:

    We propose a quantization scheme based on a Kd (K dimensional)-tree algorithm for signature sequence in a reverse-link direct sequence (DS)-code division multiple access (CDMA). With a few feedback bits, a receiver quantizes the optimal signature that Minimizes Interference for a desired user, and relays it to the user via an error-free feedback channel. A user performance depends on the quantization codebook and a number of feedback bits. We show the performance of the proposed Kd-tree codebook with a nearest neighbor criterion and derive the performance approximation. Also we modify the Kd-tree scheme to search for the entry in the codebook, which gives the least Interference. Numerical examples show that the Kd-tree codebook performs close to the optimal codebook with only fraction of computational complexity.

S Mclaughlin - One of the best experts on this subject based on the ideXlab platform.

  • congestion based routing strategies in multihop tdd cdma networks
    IEEE Journal on Selected Areas in Communications, 2005
    Co-Authors: T Rouse, S Mclaughlin
    Abstract:

    In this paper, a network topology is investigated that allows both peer-to-peer and nonlocal traffic in a cellular-based time-division duplex code-division multiple-access (TDD-CDMA) system known as opportunity driven multiple access (ODMA). The key to offering appropriate performance of peer-to-peer communication in such a system relies on the use of a routing algorithm which Minimizes Interference. This paper presents a study of the constraints and limitations on the capacity of such a system using a variety of routing techniques. A congestion-based routing algorithm is presented that attempts to minimize the overall power of the system as well as providing a measure of feasibility. This technique provides the lowest required transmit power in all circumstances, and the highest capacity in nearly all cases studied. All of the routing algorithms studied here allocate TDD time slots on a first come first served basis according to a set of predefined rules. This fact is utilized to enable the development of a combined routing and resource allocation algorithm for TDD-CDMA relaying. A novel method of time slot allocation according to relaying requirements is then developed. Two measures of assessing congestion are presented based on matrix norms. One is suitable for current interior point solution, the other is more elegant but is not currently suitable for efficient minimization and, thus, practical implementation.

Chuan Wu - One of the best experts on this subject based on the ideXlab platform.

  • Deep Learning-based Job Placement in Distributed Machine Learning Clusters
    IEEE INFOCOM 2019 - IEEE Conference on Computer Communications, 2019
    Co-Authors: Yanghua Peng, Chuan Wu
    Abstract:

    Production machine learning (ML) clusters commonly host a variety of distributed ML workloads, e.g., speech recognition, machine translation. While server sharing among jobs improves resource utilization, Interference among co-located ML jobs can lead to significant performance downgrade. Existing cluster schedulers (e.g., Mesos) are Interference-oblivious in their job placement, causing suboptimal resource efficiency. Interference-aware job placement has been studied in the literature, but was treated using detailed workload profiling and Interference modeling, which is not a general solution. This paper presents Harmony, a deep learning-driven ML cluster scheduler that places training jobs in a manner that Minimizes Interference and maximizes performance (i.e., training completion time). Harmony is based on a carefully designed deep reinforcement learning (DRL) framework augmented with reward modeling. The DRL employs state-of-the-art techniques to stabilize training and improve convergence, including actor-critic algorithm, job-aware action space exploration and experience replay. In view of a common lack of reward samples corresponding to different placement decisions, we build an auxiliary reward prediction model, which is trained using historical samples and used for producing reward for unseen placement. Experiments using real ML workloads in a Kubernetes cluster of 6 GPU servers show that Harmony outperforms representative schedulers by 25% in terms of average job completion time.

Laurent Decreusefond - One of the best experts on this subject based on the ideXlab platform.

  • 3D Beamforming based Dynamic TDD Interference Mitigation Scheme
    2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 2020
    Co-Authors: Jalal Rachad, Ridha Nasri, Laurent Decreusefond
    Abstract:

    Dynamic Time Division Duplexing (D-TDD) allows cells to accommodate asymmetric traffic variations with high resource assignment flexibility. However, this feature is limited by two additional types of Interference between cells in opposite transmission direction: downlink (DL) to uplink (UL) and UL to DL Interference. Therefore, using this mode with macrocell deployments requires Interference mitigation techniques to reduce the strong DL to UL Interference. 3D beamforming is an efficient technique that Minimizes Interference and enhances performance by exploiting a large 2D array of antennas intelligently. Combining D-TDD and 3D beamforming can make DTDD feasible for macro-cells. The aim of this work is to provide a 3D beamforming analytical model in a D-TDD based macro-cells' deployment where beamforming horizontal and vertical radiation patterns depend on the spatial distribution of random users' locations. We evaluate Interference in terms of Interference to Signal Ratio (ISR). We show that the cumulative ISR can be written in terms of convergent series and its expectation is an almost sure convergent series. Different numerical results are presented to justify the applicability of this scheme.

David Wool - One of the best experts on this subject based on the ideXlab platform.

  • phloem feeding specialists sharing a host tree resource partitioning Minimizes Interference competition among galling aphid species
    Oikos, 1995
    Co-Authors: Moshe Inbar, David Wool
    Abstract:

    This study deals with a guild of specialist, phloem feeding aphids (Homoptera; Pemphigidae; Fordinae) which form galls on Pistachia trees (Anacardiaceae). In Israel, each of two Pistacia species carries five common species of galling aphids, which may occur in large numbers on the same individual tree, shoot, leaf or even leaflet. These Fordinae have very similar life cycles, they all feed on phloem sap, and all of them need very young, unfolding leaves for gall formation. Our data show, however, that most pairs of species are unlikely to compete for galling sites because their niches are separated either temporally (fundatrices arrive at different times and, therefore, occupy different leaves) or spatially, by attacking different sites on the tree, shoot, or leaf. In 1991-1993, we calculated niche breadth (B) of each species, and proportional similarity (PS) between each species pair on shoots colonized by more than one species. PS between different species pairs on leaves within shoots ranged between 0 and 0.48. This means that, on the niche dimension represented by the shoot, the two species may share some of the habitat units. If the two species compete, we expected that niche breadth of at least one of them would be reduced on the cohabited shoots. This was not the case. B was not negatively affected when pairs of species occupied the same shoot, compared with shoots where only one species was present. B was positively correlated with density: the chance of niche overlap increased when more galls occupied the same shoot. In species sharing leaves within shoots, B showed similar trends. On P. atlantica, Smynthurodes betae and Forda riccobonii had the highest PS. Both the species make their galls on leaflet margins and occur frequently on the same leaves. Only at this level did we detect negative interactions: the frequency of co-occurrence of galls of both species on the same leaflet (within galled leaves) was significantly less than expected by chance. Taking into account our deliberate non-random selection of trees, shoots, and leaves, where competition was likely to occur, we conclude that interspecific Interference competition for galling sites is not a major determinant of the resource partitioning of these closely-related herbivores