-
作者：Singh, Navjot; Data, Deepesh; George, Jemin; Diggavi, Suhas
作者单位：University of California System; University of California Los Angeles; United States Department of Defense; United States Army; US Army Research, Development & Engineering Command (RDECOM); US Army Research Laboratory (ARL)
摘要：In this article, we propose and analyze SParsified Action Regulated Quantized-Stochastic Gradient Descent (SPARQ-SGD), a communication-efficient algorithm for decentralized training of large-scale machine learning models over a graph with $n$ nodes, where communication efficiency is achieved using compressed exchange of local model parameters among neighboring nodes, which is triggered only when an event (a locally computable condition) is satisfied. Specifically, in SPARQ-SGD, each node takes...
-
作者：Cai, Mingyu; Xiao, Shaoping; Li, Zhijun; Kan, Zhen
作者单位：Lehigh University; University of Iowa; Chinese Academy of Sciences; University of Science & Technology of China, CAS
摘要：This paper studies optimal motion planning subject to motion and environment uncertainties. By modeling the system as a probabilistic labeled Markov decision process (PL-MDP), the control objective is to synthesize a finite-memory policy, under which the agent satisfies complex high-level tasks expressed as linear temporal logic (LTL) with desired satisfaction probability. In particular, the cost optimization of the trajectory that satisfies infinite horizon tasks is considered, and the trade-...
-
作者：Massambone, Rafael; Costa, Eduardo Fontoura; Helou, Elias Salomao
作者单位：Universidade de Sao Paulo
摘要：In this article, a stochastic incremental subgradient algorithm for the minimization of a sum of convex functions is introduced. The method sequentially uses partial subgradient information, and the sequence of partial subgradients is determined by a general Markov chain. This makes it suitable to be used in networks, where the path of information flow is stochastically selected. We prove convergence of the algorithm to a weighted objective function, where the weights are given by the Cesaro l...
-
作者：McInerney, Ian; Kerrigan, Eric C.; Constantinides, George A.
作者单位：Imperial College London; Imperial College London
摘要：First-order optimizationsolvers, such as the fast gradient method (FGM), are increasingly being used to solve model predictive control problems in resource-constrained environments. Unfortunately, the convergence rate of these solvers is significantly affected by the conditioning of the problem data, with ill-conditioned problems requiring a large number of iterations. To reduce the number of iterations required, we present a simple method for computing a horizon-independent preconditioning ma...
-
作者：Pereira, Renan L.; de Oliveira, Matheus S.
作者单位：Comando-Geral de Tecnologia Aeroespacial (CTA); Instituto Tecnologico de Aeronautica (ITA)
摘要：This technical article presents novel robust stabilization conditions for discrete-time linear parameter-varying (LPV) systems with linear fractional representation (LFR). The proposed conditions rely on the use of slack variables and decision matrices associated with the LFR approach to provide new controller designs. In addition, we address parameter-dependent Lyapunov functions and full-block multipliers to obtain less conservative synthesis conditions for discrete-time LPV/LFR systems. Des...
-
作者：Colla, Sebastien; Hendrickx, Julien M.
作者单位：Universite Catholique Louvain
摘要：In this article, we present a methodology to automatically compute worst-case performance bounds for a large class of first-order decentralized optimization algorithms. These algorithms aim at minimizing the average of local functions that are distributed across a network of agents. They typically combine local computations and consensus steps. Our methodology is based on the approach of performance estimation problem (PEP), which allows computing the worst-case performance and a worst-case in...
-
作者：Kim, Dohyun
作者单位：Sungkyunkwan University (SKKU)
摘要：We establish the convergence of a heterogeneous multiagent system on the unit sphere in even-dimensional Euclidean space. In this work, we show that all relative distances converge to some definite values, when the natural frequencies are completely determined diagonal matrices with half-dimension and the coupling strength is sufficiently larger than the maximal differences of the natural frequencies. Our result extends the previous results about practical aggregation where the maximal distanc...
-
作者：Sau, Nguyen Huu; Thuan, Mai Viet
作者单位：Hanoi University of Industry (HaUI)
摘要：This article deals with the impulsive stability problem for positive linear differential-algebraic equations (DAEs) with a time-varying delay. First, a sufficient condition for the positivity of the considered system is addressed. Then, a new exponential stability criterion for positive linear DAEs with a time-varying delay is obtained based on a suitable change of variables. According to our results, a stable impulse-free system can keep its original stability property under certain destabili...
-
作者：Zhang, Xiaoyu; Gao, Chuanhou; Dochain, Denis
作者单位：Zhejiang University; Zhejiang University; Universite Catholique Louvain
摘要：Motivated by the concept of the endotactic network, a kind of special geometric structure in chemical reaction networks developed for persistence analysis, we propose a new notion, named W-I-endotactic network. The corresponding network set is a larger class of set than the endotactic network, let alone the weakly reversible network. Based on an energy-like function, we prove that all 1-dimensional mass-action W-I-endotactic networks are persistent. Furthermore, we prove some higher dimensiona...
-
作者：Wang, Le Yi; Yin, George; Lin, Feng; Polis, Michael P.; Chen, Wen
作者单位：Wayne State University; University of Connecticut; Wayne State University; Oakland University; Wayne State University
摘要：This article investigates the problem of joint continuous and discrete state estimation of randomly switched linear systems in which subsystems may not be observable. Estimation of both continuous state and discrete sequence simultaneously based on the same output observations is a challenging task that is inherently nonlinear and often infinite dimensional. This article presents necessary and sufficient conditions when joint estimation is possible without using a probing input. When such cond...
