In this paper, a general nonlinear 1st-order consensus-based solution for distributed constrained convex optimization is considered for applications in network resource allocation. The proposed continuous-time solution is used to optimize continuously-differentiable strictly convex cost functions over weakly-connected undirected multi-agent networks. The solution is anytime feasible and models various nonlinearities to account for imperfections and constraints on the (physical model of) agents in terms of their limited actuation capabilities, e.g., quantization and saturation constraints among others. Moreover, different applications impose specific nonlinearities to the model, e.g., convergence in fixed/finite-time, robustness to uncertainties, and noise-tolerant dynamics. Our proposed distributed resource allocation protocol generalizes such nonlinear models. Putting convex set analysis together with the Lyapunov theorem, we provide a general technique to prove convergence (i) regardless of the particular type of nonlinearity (ii) with weak network-connectivity requirement (i.e., uniform-connectivity). We simulate the performance of the protocol in continuous-time coordination of generators, known as the economic dispatch problem (EDP).
相關內容
Federated Learning (FL) enables distributed training by learners using local data, thereby enhancing privacy and reducing communication. However, it presents numerous challenges relating to the heterogeneity of the data distribution, device capabilities, and participant availability as deployments scale, which can impact both model convergence and bias. Existing FL schemes use random participant selection to improve fairness; however, this can result in inefficient use of resources and lower quality training. In this work, we systematically address the question of resource efficiency in FL, showing the benefits of intelligent participant selection, and incorporation of updates from straggling participants. We demonstrate how these factors enable resource efficiency while also improving trained model quality.
One of the major challenges in using distributed learning to train complicated models with large data sets is to deal with stragglers effect. As a solution, coded computation has been recently proposed to efficiently add redundancy to the computation tasks. In this technique, coding is used across data sets, and computation is done over coded data, such that the results of an arbitrary subset of worker nodes with a certain size are enough to recover the final results. The major challenges with those approaches are (1) they are limited to polynomial function computations, (2) the size of the subset of servers that we need to wait for grows with the multiplication of the size of the data set and the model complexity (the degree of the polynomial), which can be prohibitively large, (3) they are not numerically stable for computation over real numbers. In this paper, we propose Berrut Approximated Coded Computing (BACC), as an alternative approach, which is not limited to polynomial function computation. In addition, the master node can approximately calculate the final results, using the outcomes of any arbitrary subset of available worker nodes. The approximation approach is proven to be numerically stable with low computational complexity. In addition, the accuracy of the approximation is established theoretically and verified by simulation results in different settings such as distributed learning problems. In particular, BACC is used to train a deep neural network on a cluster of servers, which outperforms repetitive computation (repetition coding) in terms of the rate of convergence.
The anticipated increase in the count of IoT devices in the coming years motivates the development of efficient algorithms that can help in their effective management while keeping the power consumption low. In this paper, we propose an intelligent multi-channel resource allocation algorithm for dense LoRa networks termed LoRaDRL and provide a detailed performance evaluation. Our results demonstrate that the proposed algorithm not only significantly improves LoRaWAN's packet delivery ratio (PDR) but is also able to support mobile end-devices (EDs) while ensuring lower power consumption hence increasing both the lifetime and capacity of the network.} Most previous works focus on proposing different MAC protocols for improving the network capacity, i.e., LoRaWAN, delay before transmit etc. We show that through the use of LoRaDRL, we can achieve the same efficiency with ALOHA \textcolor{black}{compared to LoRaSim, and LoRa-MAB while moving the complexity from EDs to the gateway thus making the EDs simpler and cheaper. Furthermore, we test the performance of LoRaDRL under large-scale frequency jamming attacks and show its adaptiveness to the changes in the environment. We show that LoRaDRL's output improves the performance of state-of-the-art techniques resulting in some cases an improvement of more than 500\% in terms of PDR compared to learning-based techniques.
Federated learning (FL), an emerging distributed machine learning paradigm, in conflux with edge computing is a promising area with novel applications over mobile edge devices. In FL, since mobile devices collaborate to train a model based on their own data under the coordination of a central server by sharing just the model updates, training data is maintained private. However, without the central availability of data, computing nodes need to communicate the model updates often to attain convergence. Hence, the local computation time to create local model updates along with the time taken for transmitting them to and from the server result in a delay in the overall time. Furthermore, unreliable network connections may obstruct an efficient communication of these updates. To address these, in this paper, we propose a delay-efficient FL mechanism that reduces the overall time (consisting of both the computation and communication latencies) and communication rounds required for the model to converge. Exploring the impact of various parameters contributing to delay, we seek to balance the trade-off between wireless communication (to talk) and local computation (to work). We formulate a relation with overall time as an optimization problem and demonstrate the efficacy of our approach through extensive simulations.
Spurred by the severe restrictions on mobility due to the COVID-19 pandemic, there is currently intense interest in developing the Metaverse, to offer virtual services/business online. A key enabler of such virtual service is the digital twin, i.e., a digital replication of real-world entities in the Metaverse, e.g., city twin, avatars, etc. The real-world data collected by IoT devices and sensors are key for synchronizing the two worlds. In this paper, we consider the scenario in which a group of IoT devices are employed by the Metaverse platform to collect such data on behalf of virtual service providers (VSPs). Device owners, who are self-interested, dynamically select a VSP to maximize rewards. We adopt hybrid evolutionary dynamics, in which heterogeneous device owner populations can employ different revision protocols to update their strategies. Extensive simulations demonstrate that a hybrid protocol can lead to evolutionary stable states.
We consider the problem of online allocation (matching, budgeted allocations, and assortments) of reusable resources where an adversarial sequence of resource requests is revealed over time and allocated resources are used/rented for a stochastic duration, drawn independently from known resource usage distributions. This problem is a fundamental generalization of well studied models in online matching and resource allocation. We give an algorithm that obtains the best possible competitive ratio of $(1-1/e)$ for general usage distributions and large resource capacities. At the heart of our algorithm is a new quantity that factors in the potential of reusability for each resource by (computationally) creating an asymmetry between identical units of the resource. In order to control the stochastic dependencies induced by reusability, we introduce a relaxed online algorithm that is only subject to fluid approximations of the stochastic elements in the problem. The output of this relaxed algorithm guides the overall algorithm. Finally, we establish competitive ratio guarantees by constructing a feasible solution to an LP free system of constraints. More generally, these ideas lead to a principled approach for integrating stochastic and combinatorial elements (such as reusability, customer choice, and budgeted allocations) in online resource allocation problems.
We study the problem of optimal power allocation in single-hop multi-antenna ad-hoc wireless networks. A standard technique to solve this problem involves optimizing a tri-convex function under power constraints using a block-coordinate-descent based iterative algorithm. This approach, termed WMMSE, tends to be computationally complex and time consuming. Several learning-based approaches have been proposed to speed up the power allocation process. A recent work, UWMMSE, learns an affine transformation of a WMMSE parameter in an unfolded structure to accelerate convergence. In spite of achieving promising results, its application is limited to single-antenna wireless networks. In this work, we present a UWMMSE framework for power allocation in (multiple-input multiple-output) MIMO interference networks. A major advantage of this method lies in its use of low-complexity learnable systems in which the number of parameters scales linearly with respect to the hidden layer size of embedded neural architectures and the product of the number of transmitter and receiver antennas only, fully independent of the number of transceivers in the network. We illustrate the superiority of our method through an empirical study of our approach in comparison to WMMSE and also analyze its robustness to changes in channel conditions and network size.
There is a continuous growth in demand for time sensitive applications which has shifted the cloud paradigm from a centralized computing architecture towards distributed heterogeneous computing platforms where resources located at the edge of the network are used to provide cloud-like services. This paradigm is widely known as fog computing. Virtual machines (VMs) have been widely utilized in both paradigms to enhance the network scalability, improve resource utilization, and energy efficiency. Moreover, Passive Optical Networks (PONs) are a technology suited to handling the enormous volumes of data generated in the access network due to their energy efficiency and large bandwidth. In this paper, we utilize a PON to provide the connectivity between multiple distributed fog units to achieve federated (i.e. cooperative) computing units in the access network to serve intensive demands. We propose a mixed integer linear program (MILP) to optimize the VM placement in the federated fog computing units with the objective of minimizing the total power consumption while considering inter-VM traffic. The results show a significant power saving as a result of the proposed optimization model by up to 52%, in the VM-allocation compared to a baseline approach that allocates the VM requests while neglecting the power consumption and inter-VMs traffic in the optimization framework.
Residual networks (ResNets) have displayed impressive results in pattern recognition and, recently, have garnered considerable theoretical interest due to a perceived link with neural ordinary differential equations (neural ODEs). This link relies on the convergence of network weights to a smooth function as the number of layers increases. We investigate the properties of weights trained by stochastic gradient descent and their scaling with network depth through detailed numerical experiments. We observe the existence of scaling regimes markedly different from those assumed in neural ODE literature. Depending on certain features of the network architecture, such as the smoothness of the activation function, one may obtain an alternative ODE limit, a stochastic differential equation or neither of these. These findings cast doubts on the validity of the neural ODE model as an adequate asymptotic description of deep ResNets and point to an alternative class of differential equations as a better description of the deep network limit.
The training of deep residual neural networks (ResNets) with backpropagation has a memory cost that increases linearly with respect to the depth of the network. A way to circumvent this issue is to use reversible architectures. In this paper, we propose to change the forward rule of a ResNet by adding a momentum term. The resulting networks, momentum residual neural networks (Momentum ResNets), are invertible. Unlike previous invertible architectures, they can be used as a drop-in replacement for any existing ResNet block. We show that Momentum ResNets can be interpreted in the infinitesimal step size regime as second-order ordinary differential equations (ODEs) and exactly characterize how adding momentum progressively increases the representation capabilities of Momentum ResNets. Our analysis reveals that Momentum ResNets can learn any linear mapping up to a multiplicative factor, while ResNets cannot. In a learning to optimize setting, where convergence to a fixed point is required, we show theoretically and empirically that our method succeeds while existing invertible architectures fail. We show on CIFAR and ImageNet that Momentum ResNets have the same accuracy as ResNets, while having a much smaller memory footprint, and show that pre-trained Momentum ResNets are promising for fine-tuning models.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191