Showing papers by "Richard H. Middleton published in 2021"

Straggler Effect Mitigation for Federated Learning in Cell-Free Massive MIMO

Abstract: Straggler effect is the main bottleneck in realizing federated learning (FL) in wireless networks. This work proposes a novel user (UE) selection approach to mitigate this effect with UE sampling in cell-free massive multiple-input multiple-output networks. Our proposed approach selects only a small subset of UEs for participating in one FL process. Importantly, since the UEs are selected before any FL process is executed, the performance of FL during the executing time is not affected by our method. Here, we select UEs by solving an FL transmission time minimization problem that jointly optimizes UE selection, power control, and data rate. The problem is formulated to capture the complex interactions among the FL training time, UE selection, and straggler effect. This mixed-integer mixed-timescale stochastic nonconvex problem is constrained by the minimum number of UEs to guarantee the quality of learning. By employing online successive convex approximation, we propose a novel algorithm to solve the formulated problem with guaranteed convergence to the neighbourhood of their stationary points. Our approach can significantly reduce the FL transmission time over baseline approaches, especially in the networks that experience serious straggler effect due to the moderately low density of access points.

Scalability in Nonlinear Network Systems Affected by Delays and Disturbances

Abstract: This article is concerned with the study of scalability in nonlinear heterogeneous networks affected by communication delays and disturbances. After formalizing the notion of scalability, we give two sufficient conditions to assess this property. Our results can be used to study leader–follower and leaderless networks and allow us to consider the case when the desired configuration of the system changes over time. We show how our conditions can be turned into design guidelines to guarantee scalability and illustrate their effectiveness via numerical examples.

Autonomous Synchronization of Heterogeneous Multiagent Systems

Abstract: This article formulates a new type of synchronization problem for heterogeneous multiagent systems (MASs), called autonomous synchronization. In this problem, neither the synchronized agent dynamics nor the synchronized states are specified a priori , instead, they are autonomously determined by the inherent properties and the initial states of agents, thus providing an MAS with more degrees of adaptability and higher synchronization efficiency. To achieve autonomous synchronization, a novel dynamics update law and a synchronizing control law are proposed and the sufficient solvability conditions are explicitly revealed. The results are supported by rigorous theoretical analysis and numerical simulation, in both continuous-time and discrete-time settings.

Understanding the dynamics of inter-provincial migration in the Mekong Delta, Vietnam: an agent-based modeling study:

Abstract: Recent large-scale migration flows from rural areas of the Mekong Delta (MKD) to larger cities in the South-East (SE) region of Vietnam have created the largest migration corridor in the country. T...

Benders decomposition for a reverse logistics network design problem in the dairy industry

Abstract: Designing a value-creating whey recovery network is an important reverse logistics problem in the dairy industry. Whey is a byproduct of cheese making with many potential applications. Due to environmental legislation and economic advantages, raw whey should be processed into commercial products rather than disposed of into the environment. In this paper, we study a whey reverse logistics network design problem under demand uncertainty, where demand is the amount of raw whey produced by a set of cheese makers. We formulate the problem as a hierarchical facility location problem with two levels of facilities and use two-stage stochastic programming to tackle the issue of uncertainty. We consider a sample average approximation method to estimate the expected cost and employ an accelerated Benders decomposition algorithm to solve the resulting formulation to optimality. An extensive computational study, using 1200 benchmark instances of the problem, demonstrates the efficacy of our improved algorithm. Instances with as many as 20 cheese makers are shown to be solved by our proposed methodology an order of magnitude faster than the automatic Benders decomposition algorithm offered by a commercial solver. Optimal solutions of a real case study with 51 cheese makers together with useful managerial insights are also reported. The value of stochastic solution in the case study signifies the importance of considering the uncertainties that are inherent in the dairy industry. Our analysis of the case study shows that the total expected cost is increased by 28% if such uncertainties are ignored. Furthermore, this increase can become arbitrarily large as the outsourcing costs increase.

Discrete-Time Consensus Networks: Scalability, Grounding and Countermeasures.

Abstract: We investigate the disruption of discrete-time consensus problems via grounding. Loosely speaking, grounding a network occurs if the state of one agent no longer responds to inputs from other agents and/or changes its dynamics. Then, the agent becomes a leader or a so-called stubborn agent. The disruption of the agent can be caused by internal faults, safety protocols or due to an external malicious attack. In this paper we investigate how grounding affects expander graph families that usually exhibit good scaling properties with increasing network size. It is shown that the algebraic connectivity and eigenratio of the network decrease due to the grounding causing the performance and scalability of the network to deteriorate, even to the point of losing consensusability. We then present possible countermeasures to such disruptions and discuss their practicality and limitations. In particular, for a specific countermeasure of deliberately grounding additional nodes, we investigate extensively how to select additional nodes to ground and how many nodes we need to ground to recover the consensus performance. Our findings are supported by a wide range of numerical simulations.