Local Computation Research Articles

F3KM: Federated, Fair, and Fast k-means

This paper proposes a federated, fair, and fast k-means algorithm (F3KM) to solve the fair clustering problem efficiently in scenarios where data cannot be shared among different parties. The proposed algorithm decomposes the fair k-means problem into multiple subproblems and assigns each subproblem to a client for local computation. Our algorithm allows each client to possess multiple sensitive attributes (or have no sensitive attributes). We propose an in-processing method that employs the alternating direction method of multipliers (ADMM) to solve each subproblem. During the procedure of solving subproblems, only the computation results are exchanged between the server and the clients, without exchanging the raw data. Our theoretical analysis shows that F3KM is efficient in terms of both communication and computation complexities. Specifically, it achieves a better trade-off between utility and communication complexity, and reduces the computation complexity to linear with respect to the dataset size. Our experiments show that F3KM achieves a better trade-off between utility and fairness than other methods. Moreover, F3KM is able to cluster five million points in one hour, highlighting its impressive efficiency.

Integrated correlators with a Wilson line in N\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\mathcal{N} $$\\end{document} = 4 SYM

In the context of integrated correlators in N\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\mathcal{N} $$\\end{document} = 4 SYM, we study the 2-point functions of local operators with a superconformal line defect. Starting from the mass-deformed N\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\mathcal{N} $$\\end{document} = 2* theory in presence of a 12\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\frac{1}{2} $$\\end{document}-BPS Wilson line, we exploit the residual superconformal symmetry after the defect insertion, and show that the massive deformation corresponds to integrated insertions of the superconformal primaries belonging to the stress tensor multiplet with a specific integration measure which is explicitly derived after enforcing the superconformal Ward identities. Finally, we show how the Wilson line integrated correlator can be computed by the N\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\mathcal{N} $$\\end{document} = 2* Wilson loop vacuum expectation value on a 4-sphere in terms of a matrix model using supersymmetric localization. In particular, we reformulate previous matrix model computations by making use of recursion relations and Bessel kernels, providing a direct link with more general localization computations in N\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$ \\mathcal{N} $$\\end{document} = 2 theories.

Design and construction of an Ultrasounographic Device Dedicated to the Parietal Rheology Measurement in Human Carotid Artery

We present in this article a detection system of ultrasounographic signal (USG) dedicated to the measurement of parietal rheology in human carotid artery. The aim of this practical work is to develop an ultrasonic method capable of visualizing the localized velocity profile of the carotid artery wall. The device produced makes it possible to take the USG signal from carotid wall of the human neck. It comprises an ultrasonic contact probe applied during the measurement (sensor), a signal shaping circuit, an acquisition and digitization interface in conjunction with an RS232 serial communication port and a local computer terminal hosting a graphical user interface (GUI) implemented in Visual Basic integrated development environment enabling the display, archiving and the digital processing of parietal ultrasonic signal. The aim of this study is to reconnoitering the possibility of using ultrasound for a continuous and non-invasive measurement of the arteries parietal rheology. In this work, a sensor of ultrasound is used to measure the movement and the speed of the carotid artery wall, an image of the arteries parietal rheology.

Optimal Intervention in Transportation Networks

We study a network design problem (NDP) where the planner aims at selecting the optimal single-link intervention on a transportation network to minimize the travel time under Wardrop equilibrium flows. Our first result is that, if the delay functions are affine and the support of the equilibrium is not modified with interventions, the NDP may be formulated in terms of electrical quantities computed on a related resistor network. In particular, we show that the travel time variation corresponding to an intervention on a given link depends on the effective resistance between the endpoints of the link. We suggest an approach to approximate such an effective resistance by performing only local computation, and exploit it to design an efficient algorithm to solve the NDP. We discuss the optimality of this procedure in the limit of infinitely large networks, and provide a sufficient condition for its optimality. We then provide numerical simulations, showing that our algorithm achieves good performance even if the equilibrium support varies and the delay functions are non-linear.

Automatic Performance Estimation for Decentralized Optimization

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 instance of first-order optimization algorithms by solving an SDP. We propose two ways of representing consensus steps in PEPs, which allow writing and solving PEPs for decentralized optimization. The first formulation is exact but specific to a given averaging matrix. The second formulation is a relaxation but provides guarantees valid over an entire class of averaging matrices, characterized by their spectral range. This formulation often allows recovering a posteriori the worst possible averaging matrix for the given algorithm. We apply our methodology to three different decentralized methods. For each of them, we obtain numerically tight worst-case performance bounds that significantly improve on the existing ones, as well as insights about the parameters tuning and the worst communication networks.

Joint Optimization of Security Strength and Resource Allocation for Computation Offloading in Vehicular Edge Computing

Vehicular Edge Computing (VEC) is a promising new paradigm that has attracted much attention in recent years, which can enhance the storage and computing capabilities of vehicular networks to provide users with low latency and high-quality services. Due to the open access and unreliable wireless channels, some appropriate security measures should be implemented in the VEC to ensure information security. However, the operation of the security mechanism dominates supererogatory computing resources, thus affecting the performance of VEC systems. The scarcity of computation and energy resources of the vehicles conflicts with the requirement of tasks for time delay and information security. In this paper, taking the driving velocity and position of the vehicles, the number of lanes, the model and density of the attackers, and security strength into consideration, we formulate a max-min optimization problem to jointly optimize offloading decision, transmit power, task computation frequency, encryption computation frequency, edge computation frequency, and block length to obtain optimal secure information capacity and local computation delay. The formulated optimization problem is a mixed integer nonlinear programming (MINLP), which is intractable. We apply the generalized benders decomposition (GBD)-based method to solve it. The simulation results show that our proposed algorithms have convergence and effectiveness and achieve fairness among vehicles on the road.

APPROACHES TO INCREASE THE EFFICIENCY OF TRAINING AND TESTING BY CREATING A UNIFORM INFORMATION AND REFERENCE SYSTEM AND AUTOMATING THE ACTIVITIES OF THE INSTITUTE’S SCIENTISTS

The article is devoted to ways of solving certain problematic aspects of the preparation and conducting of tests, caused by the lack of a centralized approach to the creation, replenishment and access of scientists of the institute to the regulatory and legal framework and other materials necessary for the development of organizational and methodical documents for conducting tests. This causes long-term, not always high-quality preparation and testing of weapons, military, special equipment, ammunition, software and other types of experimental samples.
 The authors of the article outline approaches to increase the efficiency, automation, and quality of preparation and testing. For this purpose, it is proposed to create a single information and reference system at the institute, built on the basis of global (cloud environment) and local (local computer network) data exchange systems.
 The proposed approaches provide for the implementation of advanced information technologies that ensure the accumulation, search, access and use by scientists of the State Research Institute of Testing and Certification of Weapons and Military Equipment of the necessary regulatory and legal acts, organizational and methodological documents, reporting materials and other reference data at the stages preparation and testing of weapons, military, special equipment, ammunition, software and other types of prototypes.
 The introduction of the proposed approaches will reduce the cost of human resources for the preparation for the tests, will ensure the improvement of the quality of working out the organizational and methodical documents, which will ensure the proper conduct of the tests by the test team.

Optimal Pricing Design for Coordinated and Uncoordinated IoT Networks

An Internet of Things (IoT) system can include several different types of service providers, who sell IoT service, network service, and computation service to customers, either jointly or separately. A deep understanding of complicated coupling among these providers in terms of pricing and service decisions is critical to the success of IoT networks. This paper studies the impact of the provider interaction structures on the overall IoT system with heterogeneous customers. Specifically, we first study a generic IoT scenario with three interaction structures: coordinated, vertically-uncoordinated, and horizontally-uncoordinated structures. Despite the challenging non-convex optimization problems involved in modeling and analyzing these structures, we successfully obtain the closed-form optimal pricing strategies of providers in each interaction structure. We further extend the analysis to a specific IoT scenario with local computation capability (e.g., Internet of Vehicles (IoV)). We prove that the coordinated structure is better than two uncoordinated structures for both providers and customers, as it avoids selfish price markup behaviors in uncoordinated structures. Between the two uncoordinated structures, when customers' demand variance is large and utility-cost ratio is medium, vertically-uncoordinated structure is better than horizontal one for both providers and customers, due to the complementary providers' competition in horizontally-uncoordinated structure. Counter-intuitively, we identify that providers' optimal prices do not change with their costs at the critical point of customers' full participation in the vertically-uncoordinated structure.

On the partition noise in chosen particle weighted methods and its consequences for weakly-compressible flow models

The Meshless Finite Volume (MFV) method, which is a member of the particle weighted methods (PWM) family, is analysed in the context of the so-called partition noise problem. This issue is inherent to PWM-type discretisation which is based on compact support kernel functions and particle sampling. The partition noise exhibits itself as spurious oscillations observed in the solution; they are caused by unphysical source terms in the discretised equations and approximate local volumes computation. Unfortunately, decreasing the discrete length scale which controls the spatial resolution reduces the partition noise only weakly (and never completely). When choosing numerical schemes, a special care is required since in naive implementations the partition noise may even lead to loss of stability. However, even when stability is assured, PWM does not fulfil the assumptions of the Lax equivalence theorem (i.e. the discretised equations are not consistent), therefore convergence in the classical meaning cannot be guaranteed.As an example of a problematic use of MFV, we investigate its application to the solution of fluid flow equations at low Mach numbers, aiming to mimic truly incompressible flow conditions. Taken here as isothermal Euler equations, they are the basis of the so-called weakly-compressible (WC) flow models. Such modelling does not involve any elliptic-type equation and, due to the algorithmic and computational reasons, seems to be well-suited for particle methods. The WC flow model imposes, however, a serious challenge for PWM since: (i) the method is not gauge invariant and the disparity between the speed of sound and the convective velocity scale causes a significant spurious source of momentum; (ii) the inaccuracy in the local volumes approximation leads to errors in the computed density which are comparable to physically relevant slight variations, inherent to the WC approach. We propose specific numerical and modelling measures: controlled relaxation of initial particle distribution, particle volumes computation using an optimal kernel function, a choice of gradient limiting techniques and a carefully chosen background pressure. Although these improvements significantly increase the accuracy (however at remarkably increased computational cost), the theoretical 2nd-order convergence is achieved only at lower spatial resolutions when the truncation errors dominate. The accuracy of mesh-based methods is achieved in PWM only for the pressure field; the errors in the velocity field are much larger and convergence is at most 1st-order. We point some contradictory requirements of the model equations and the numerical method. One of the conclusions is that it is doubtful to construct a convergent truly Lagrangean method of PWM type at reasonable computational cost. The present findings also apply to conservative variants of the weakly-compressible Smoothed Particle Hydrodynamics which, being first order consistent in the best case, suffers from the partition noise even more severely than the considered MFV approach which is theoretically second order consistent. The conclusions pose a question about a rationale for the use of WC-PWM approach in chosen applications, in particular free surface and interfacial flows.

Advancing Federated Learning through Verifiable Computations and Homomorphic Encryption.

Federated learning, as one of the three main technical routes for privacy computing, has been widely studied and applied in both academia and industry. However, malicious nodes may tamper with the algorithm execution process or submit false learning results, which directly affects the performance of federated learning. In addition, learning nodes can easily obtain the global model. In practical applications, we would like to obtain the federated learning results only by the demand side. Unfortunately, no discussion on protecting the privacy of the global model is found in the existing research. As emerging cryptographic tools, the zero-knowledge virtual machine (ZKVM) and homomorphic encryption provide new ideas for the design of federated learning frameworks. We have introduced ZKVM for the first time, creating learning nodes as local computing provers. This provides execution integrity proofs for multi-class machine learning algorithms. Meanwhile, we discuss how to generate verifiable proofs for large-scale machine learning tasks under resource constraints. In addition, we implement the fully homomorphic encryption (FHE) scheme in ZKVM. We encrypt the model weights so that the federated learning nodes always collaborate in the ciphertext space. The real results can be obtained only after the demand side decrypts them using the private key. The innovativeness of this paper is demonstrated in the following aspects: 1. We introduce the ZKVM for the first time, which achieves zero-knowledge proofs (ZKP) for machine learning tasks with multiple classes and arbitrary scales. 2. We encrypt the global model, which protects the model privacy during local computation and transmission. 3. We propose and implement a new federated learning framework. We measure the verification costs under different federated learning rounds on the IRIS dataset. Despite the impact of homomorphic encryption on computational accuracy, the framework proposed in this paper achieves a satisfactory 90% model accuracy. Our framework is highly secure and is expected to further improve the overall efficiency as cryptographic tools continue to evolve.

Dynamic multi‐agent dc‐bus reconfiguration in modular motor drives with a stacked polyphase bridge converter

AbstractA stacked polyphase bridge converter consists of multiple standard, three‐phase two‐level voltage source inverters connected in series to the same dc voltage source. The authors present a decentralised and dynamic dc‐bus reconfiguration strategy for this type of converter. The innovation of the proposed strategy lies in its ability to isolate or reactivate one of the inverters in the series connection online, even in motoring mode, enabling increased fault tolerance and an extended speed range. Both the required hardware alterations and the multi‐agent control strategy are addressed. Its major benefits compared to the state‐of‐the‐art are that only two additional active, unidirectional semiconductor switches are required per inverter, and that the control is based on local computations, local measurements and neighbour‐to‐neighbour communication only. Hence, the scalability and reliability of the hardware are extended towards the control. Simulations and experimental results on a 4 kW modular axial‐flux PMSM prove the feasibility of the concept, and validate that the dc‐bus can be reconfigured in 100 ms, without torque interruption.

Pipelines for automating compliance-based elimination and extension (PACE2): a systematic framework for high-throughput biomolecular materials simulation workflows

The formation of biomolecular materials via dynamical interfacial processes, such as self-assembly and fusion, for diverse compositions and external conditions can be efficiently probed using ensemble Molecular Dynamics (MD). However, this approach requires many simulations when investigating a large composition phase space. In addition, there is difficulty in predicting whether each simulation will yield biomolecular materials with the desired properties or outcomes and how long each simulation will run. These difficulties can be overcome by rules-based management systems, including intermittent inspection, variable sampling, and premature termination or extension of the individual MD simulations. Automating such a management system can significantly improve runtime efficiency and reduce the burden of organizing large ensembles of MD simulations. To this end, a computational framework, the Pipelines for Automating Compliance-based Elimination and Extension (PACE2), is proposed for high-throughput ensemble biomolecular materials simulations. The PACE2 framework encompasses Candidate pipelines, where each pipeline includes temporally separated simulation and analysis tasks. When a MD simulation is completed, an analysis task is triggered, which evaluates the MD trajectory for compliance. Compliant simulations are extended to the next MD phase with a suitable sample rate to allow additional, detailed analysis. Non-compliant simulations are eliminated, and their computational resources are reallocated or released. The framework is designed to run on local desktop computers and high-performance computing resources. Preliminary scientific results enabled by the use of PACE2 framework are presented, which demonstrate its potential and validates its function. In the future, the framework will be extended to address generalized workflows and investigate composition-structure-property relations for other classes of materials.

Ranking influential nodes in complex network using edge weight degree based shell decomposition

Identifying critical nodes in a complex network or ranking the nodes based on their influence over the network has many utilities. As an example, these come in handy while choosing customers for viral marketing of a new product or identifying users whom to block for preventing the spreading of misinformation/rumor among others. With proper network modeling, a large number of real-world problems could be analyzed using such node ranking methods. Ranking the nodes based on their spreading influence in a complex network is an important research problem. Network topology-based techniques play a significant role in various applications that require knowing the central entities in a modeled system. Classical network centralities like degree, betweenness, and closeness followed by core decomposition techniques are applied successfully in ranking nodes. Each of them having certain limitations like purely local measure or computation heavy etc. Due to the inherent nature of the real-world problems/events, when they are modeled as a network, most of them transform into a weighted network. Even then a large number of literature is focused on unweighted networks due to the simplicity of computations and evaluation metrics. Many node ranking methods developed for unweighted networks may be employed to weighted networks but they will not be utilize the extra information available as connection strengths. Recently, many authors have started to extend the node ranking heuristics for the weighted networks. In this paper, we propose an edge weight degree neighborhood (EwDN) based shell decomposition technique to rank the nodes of a weighted network. It considers both node degree and edge weight with a proper balance of importance. The SIR epidemic model has been used as a benchmark for simulating the spreading process. The number of generated shells, number of core–shell nodes, and monotonicity have been computed as performance metrics during the evaluation of the methods. Also, Kendall’s rank correlation with the benchmark SIR model for various competing heuristics is compared. A performance comparison with recent relevant methods shows a better outcome with the proposed method on real-world weighted networks.

Deep reinforcement learning-based online task offloading in mobile edge computing networks

Mobile edge computing has been a promising paradigm to reduce the computation delay of tasks and extend the battery life of Internet of Things devices. Most existing task offloading methods in mobile edge computing networks are based on stable task arrivals and homogeneous users. However, in practical network environments, the frequency at which an Internet of Things device generates tasks is typically time-variant, and user requirements for security guarantees and response performance are always differential. Motivated by this, we investigate online task offloading in mobile edge computing networks with diverse users. By dividing edge nodes into public and private ones, we present a novel collaboration architecture. Considering correlated arrivals, we construct a system model comprising local computing and cloud-edge computing models. Under a matrix-geometric solution framework, we analyze the system model to derive the overall task latency. For mobile edge computing networks with variable task arrival intensity, we propose a deep reinforcement learning-based online task scheduling algorithm to realize online task offloading. Experimental results demonstrate that compared to the offline task scheduling algorithm, for the correlation coefficient of 0, 0.1, and 0.2, the proposed online task scheduling algorithm can reduce the overall task latency by about 1.09%, 2.65%, and 4.34%, respectively.

BioEdge: Accelerating Object Detection in Bioimages with Edge-Based Distributed Inference

Convolutional neural networks (CNNs) have enabled effective object detection tasks in bioimages. Unfortunately, implementing such an object detection model can be computationally intensive, especially on resource-limited hardware in a laboratory or hospital setting. This study aims to develop a framework called BioEdge that can accelerate object detection using Scaled-YOLOv4 and YOLOv7 by leveraging edge computing for bioimage analysis. BioEdge employs a distributed inference technique with Scaled-YOLOv4 and YOLOv7 to harness the computational resources of both a local computer and an edge server, enabling rapid detection of COVID-19 abnormalities in chest radiographs. By implementing distributed inference techniques, BioEdge addresses privacy concerns that can arise when transmitting biomedical data to an edge server. Additionally, it incorporates a computationally lightweight autoencoder at the split point to reduce data transmission overhead. For evaluation, this study utilizes the COVID-19 dataset provided by the Society for Imaging Informatics in Medicine (SIIM). BioEdge is shown to improve the inference latency of Scaled-YOLOv4 and YOLOv7 by up to 6.28 times with negligible accuracy loss compared to local computer execution in our evaluation setting.

Efficient path planning for AUVs in unmapped marine environments using a hybrid local–global strategy

The ability of autonomous undersea vehicles (AUVs) to plan paths in unknown marine environments is the precondition for executing complicated missions. However, existing path planning algorithms based on underwater sensing equipment often struggle to achieve efficient exploration and generate high-quality trajectories. In this paper, we introduce a novel approach to efficiently handle the challenge of AUV navigation under limited information. Our solution combines global and local planning techniques to generate optimized paths that guarantee collision-free and efficient operations. In global path planning, we incrementally use the rolling windows to make decisions on high-level path branching while utilizing waypoints from selected branches to refine the calculation of local paths for enhanced accuracy. We employ an efficient small-scale path search strategy at the local path computation level by leveraging sensor-detected environments. In this stage, we propose an advanced rapidly exploring random tree (RRT) algorithm called circle-RRT. By combining adaptive circle sampling with dynamic step sizes, this algorithm can significantly reduce the generation of redundant sampling points and improve the efficiency of local path planning. We evaluated the efficiency of our algorithm in unknown environments through simulations and compared it with previous leading methods.

Joint Mobility Control and MEC Offloading for Hybrid Satellite-Terrestrial-Network-Enabled Robots

Benefiting from the fusion of communication and intelligent technologies, network-enabled robots have become important to support future machine-assisted and unmanned applications. To provide high-quality services for robots in wide areas, hybrid satellite-terrestrial networks are a key technology. Through hybrid networks, computation-intensive and latency-sensitive tasks can be offloaded to mobile edge computing (MEC) servers. However, due to the mobility of mobile robots and unreliable wireless network environments, excessive local computations and frequent service migrations may significantly increase the service delay. To address this issue, this paper aims to minimize the average task completion time for MEC-based offloading initiated by satellite-terrestrial-network-enabled robots. Different from conventional mobility-aware schemes, the proposed scheme makes the offloading decision by jointly considering the mobility control of robots. A joint optimization problem of task offloading and velocity control is formulated. Using Lyapunov optimization, the original optimization is decomposed into a velocity control subproblem and a task offloading subproblem. Then, based on the Markov decision process (MDP), a dual-agent reinforcement learning (RL) algorithm is proposed. The convergence and complexity of the improved RL algorithm are theoretically analyzed, and the simulation results show that the proposed scheme can effectively reduce the offloading delay.

NBSync: Parallelism of Local Computing and Global Synchronization for Fast Distributed Machine Learning in WANs

NBSync: Parallelism of Local Computing and Global Synchronization for Fast Distributed Machine Learning in WANs

THE RELEVANCE OF AUTOMATING THE SCHEDULE OF CLASSES IN HIGHER MILITARY EDUCATIONAL INSTITUTIONS

In the modern world, much attention is paid to the development of information technologies in the XXI century. As the President of our country noted, the formation of a national informatization system and the mass introduction of modern information and communication technologies in all spheres of society requires future specialists to have special knowledge and skills in the field of computer technology. Therefore, algorithmization and programming is of particular importance in the formation of a national informatization system, especially for the prompt resolution of important military issues and their solutions. As already noted, the system of training, advanced training and retraining of military specialists in the field of programming is one of the main tasks today, as well as the development of local computer programs designed to solve military tasks, scheduling classes, automating the educational process as a whole and putting them into practice.

Decentralized multi-task stochastic optimization with compressed communications

We consider a multi-agent network where each node has a stochastic (local) cost function that depends on the decision variable of that node and a random variable, and further, the decision variables of neighboring nodes are pairwise constrained. There is an aggregated objective function for the network, composed additively of the expected values of the local cost functions at the nodes, and the overall goal of the network is to obtain the minimizing solution to this aggregate objective function subject to all the pairwise constraints. This is to be achieved at the level of the nodes using decentralized information and local computation, with exchanges of only compressed information allowed by neighboring nodes. The paper develops algorithms and obtains performance bounds for two different models of local information availability at the nodes: (i) sample feedback, where each node has direct access to samples of the local random variable to evaluate its local cost, and (ii) bandit feedback, where samples of the random variables are not available, but only the values of the local cost functions at two random points close to the decision are available to each node. For both models, with compressed communication between neighbors, we have developed decentralized saddle-point algorithms that deliver performances no different (in order sense) from those without communication compression; specifically, we show that deviation from the global minimum value and violations of the constraints are upper-bounded by O(T−12) and O(T−14), respectively, where T is the number of iterations. Numerical examples provided in the paper corroborate these bounds.