Imperial College London

ProfessorDenizGunduz

Faculty of EngineeringDepartment of Electrical and Electronic Engineering

Professor in Information Processing
 
 
 
//

Contact

 

+44 (0)20 7594 6218d.gunduz Website

 
 
//

Assistant

 

Ms Joan O'Brien +44 (0)20 7594 6316

 
//

Location

 

1016Electrical EngineeringSouth Kensington Campus

//

Summary

 

Publications

Publication Type
Year
to

281 results found

Amiri MM, Duman TM, Gunduz D, Kulkarni SR, Poor HVet al., 2021, Blind Federated Edge Learning, IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, Vol: 20, Pages: 5129-5143, ISSN: 1536-1276

Journal article

Amiria MM, Gunduzb D, Kulkarni SR, Vincent Poor Het al., 2021, Convergence of update aware device scheduling for federated learning at the wireless edge, IEEE Transactions on Wireless Communications, Vol: 20, Pages: 3643-3658, ISSN: 1536-1276

We study federated learning (FL) at the wireless edge, where power-limited devices with local datasets collaboratively train a joint model with the help of a remote parameter server (PS). We assume that the devices are connected to the PS through a bandwidth-limited shared wireless channel. At each iteration of FL, a subset of the devices are scheduled to transmit their local model updates to the PS over orthogonal channel resources, while each participating device must compress its model update to accommodate to its link capacity. We design novel scheduling and resource allocation policies that decide on the subset of the devices to transmit at each round, and how the resources should be allocated among the participating devices, not only based on their channel conditions, but also on the significance of their local model updates. We then establish convergence of a wireless FL algorithm with device scheduling, where devices have limited capacity to convey their messages. The results of numerical experiments show that the proposed scheduling policy, based on both the channel conditions and the significance of the local model updates, provides a better long-term performance than scheduling policies based only on either of the two metrics individually. Furthermore, we observe that when the data is independent and identically distributed (i.i.d.) across devices, selecting a single device at each round provides the best performance, while when the data distribution is non-i. i.d., scheduling multiple devices at each round improves the performance. This observation is verified by the convergence result, which shows that the number of scheduled devices should increase for a less diverse and more biased data distribution.

Journal article

Ceran E, Gunduz D, Gyorgy A, 2021, A reinforcement learning approach to age of information in multi-user networks with HARQ‌, IEEE Journal on Selected Areas in Communications, Vol: 39, Pages: 1412-1426, ISSN: 0733-8716

Scheduling the transmission of time-sensitive information from a source node to multiple users over error-prone communication channels is studied with the goal of minimizing the long-term average age of information (AoI) at the users. A long-term average resource constraint is imposed on the source, which limits the average number of transmissions. The source can transmit only to a single user at each time slot, and after each transmission, it receives an instantaneous ACK/NACK feedback from the intended receiver, and decides when and to which user to transmit the next update. Assuming the channel statistics are known, the optimal scheduling policy is studied for both the standard automatic repeat request (ARQ) and hybrid ARQ (HARQ) protocols. Then, a reinforcement learning (RL) approach is introduced to find a near-optimal policy, which does not assume any a priori information on the random processes governing the channel states. Different RL methods including average-cost SARSA with linear function approximation (LFA), upper confidence reinforcement learning (UCRL2), and deep Q-network (DQN) are applied and compared through numerical simulations.

Journal article

Mashhadi MB, Yang Q, Gunduz D, 2021, Distributed deep convolutional compression for massive MIMO CSI feedback, IEEE Transactions on Wireless Communications, Vol: 20, Pages: 2621-2633, ISSN: 1536-1276

Massive multiple-input multiple-output (MIMO) systems require downlink channel state information (CSI) at the base station (BS) to achieve spatial diversity and multiplexing gains. In a frequency division duplex (FDD) multiuser massive MIMO network, each user needs to compress and feedback its downlink CSI to the BS. The CSI overhead scales with the number of antennas, users and subcarriers, and becomes a major bottleneck for the overall spectral efficiency. In this paper, we propose a deep learning (DL)-based CSI compression scheme, called DeepCMC, composed of convolutional layers followed by quantization and entropy coding blocks. In comparison with previous DL-based CSI reduction structures, DeepCMC proposes a novel fully-convolutional neural network (NN) architecture, with residual layers at the decoder, and incorporates quantization and entropy coding blocks into its design. DeepCMC is trained to minimize a weighted rate-distortion cost, which enables a trade-off between the CSI quality and its feedback overhead. Simulation results demonstrate that DeepCMC outperforms the state of the art CSI compression schemes in terms of the reconstruction quality of CSI for the same compression rate. We also propose a distributed version of DeepCMC for a multi-user MIMO scenario to encode and reconstruct the CSI from multiple users in a distributed manner. Distributed DeepCMC not only utilizes the inherent CSI structures of a single MIMO user for compression, but also benefits from the correlations among the channel matrices of nearby users to further improve the performance in comparison with DeepCMC. We also propose a reduced-complexity training method for distributed DeepCMC, allowing to scale it to multiple users, and suggest a cluster-based distributed DeepCMC approach for practical implementation.

Journal article

Rassouli B, Gunduz D, 2021, On perfect privacy, IEEE Journal on Selected Areas in Information Theory, Vol: 2, Pages: 177-191, ISSN: 2641-8770

The problem of private data disclosure is studied from an informationtheoretic perspective. Considering a pair of dependent random variables$(X,Y)$, where $X$ and $Y$ denote the private and useful data, respectively,the following problem is addressed: What is the maximum information that can berevealed about $Y$ (measured by mutual information $I(Y;U)$, in which $U$ isthe revealed data), while disclosing no information about $X$ (captured by thecondition of statistical independence, i.e., $X\independent U$, and henceforth,called \textit{perfect privacy})? We analyze the supremization of\textit{utility}, i.e., $I(Y;U)$ under the condition of perfect privacy for twoscenarios: \textit{output perturbation} and \textit{full data observation}models, which correspond to the cases where a Markov kernel, called\textit{privacy-preserving mapping}, applies to $Y$ and the pair $(X,Y)$,respectively. When both $X$ and $Y$ have a finite alphabet, the linearalgebraic analysis involved in the solution provides some interesting results,such as upper/lower bounds on the size of the released alphabet and the maximumutility. Afterwards, it is shown that for the jointly Gaussian $(X,Y)$, perfectprivacy is not possible in the output perturbation model in contrast to thefull data observation model. Finally, an asymptotic analysis is provided toobtain the rate of released information when a sufficiently small leakage isallowed. In particular, in the context of output perturbation model, it isshown that this rate is always finite when perfect privacy is not feasible, andtwo lower bounds are provided for it; When perfect privacy is feasible, it isshown that under mild conditions, this rate becomes unbounded.

Journal article

Jankowski M, Gunduz D, Mikolajczyk K, 2021, Wireless image retrieval at the edge, IEEE Journal on Selected Areas in Communications, Vol: 39, Pages: 89-100, ISSN: 0733-8716

We study the image retrieval problem at the wireless edge, where an edge device captures an image, which is then used to retrieve similar images from an edge server. These can be images of the same person or a vehicle taken from other cameras at different times and locations. Our goal is to maximize the accuracy of the retrieval task under power and bandwidth constraints over the wireless link. Due to the stringent delay constraint of the underlying application, sending the whole image at a sufficient quality is not possible. We propose two alternative schemes based on digital and analog communications, respectively. In the digital approach, we first propose a deep neural network (DNN) aided retrieval-oriented image compression scheme, whose output bit sequence is transmitted over the channel using conventional channel codes. In the analog joint source and channel coding (JSCC) approach, the feature vectors are directly mapped into channel symbols. We evaluate both schemes on image based re-identification (re-ID) tasks under different channel conditions, including both static and fading channels. We show that the JSCC scheme significantly increases the end-to-end accuracy, speeds up the encoding process, and provides graceful degradation with channel conditions. The proposed architecture is evaluated through extensive simulations on different datasets and channel conditions, as well as through ablation studies.

Journal article

Li GY, Saad W, Ozgur A, Kairouz P, Qin Z, Hoydis J, Han Z, Gunduz D, Elmirghani Jet al., 2021, Series editorial: inauguration issue of the series on machine learning in communications and networks, IEEE Journal on Selected Areas in Communications, Vol: 39, Pages: 1-3, ISSN: 0733-8716

In the era of the new generation of communication systems, data traffic is expected to continuously strain the capacity of future communication networks. Along with the remarkable growth in data traffic, new applications, such as wearable devices, autonomous systems, and the Internet of Things (IoT), continue to emerge and generate even more data traffic with vastly different requirements. This growth in the application domain brings forward an inevitable need for more intelligent processing, operation, and optimization of future communication networks.

Journal article

Gunduz D, Burth Kurka D, Jankowski M, Mohammadi Amiri M, Ozfatura ME, Sreekumar Set al., 2020, Communicate to learn at the edge, IEEE Communications Magazine, Vol: 58, Pages: 14-19, ISSN: 0163-6804

Bringing the success of modern machine learning (ML) techniques to mobile devices can enablemany new services and businesses, but also poses significant technical and research challenges. Twofactors that are critical for the success of ML algorithms are massive amounts of data and process-ing power, both of which are plentiful, yet highly distributed at the network edge. Moreover, edgedevices are connected through bandwidth- and power-limited wireless links that suffer from noise,time-variations, and interference. Information and coding theory have laid the foundations of reliableand efficient communications in the presence of channel imperfections, whose application in modernwireless networks have been a tremendous success. However, there is a clear disconnect between thecurrent coding and communication schemes, and the ML algorithms deployed at the network edge. Inthis paper, we challenge the current approach that treats these problems separately, and argue for a jointcommunication and learning paradigm for both the training and inference stages of edge learning.

Journal article

Zhu G, Du Y, Gunduz D, Huang Ket al., 2020, One-bit over-the-air aggregation for communication-efficient federated edge learning: design and convergence analysis, IEEE Transactions on Wireless Communications, Vol: 20, Pages: 2120-2135, ISSN: 1536-1276

Federated edge learning (FEEL) is a popular framework for model training at an edge server using data distributed at edge devices (e.g., smart-phones and sensors) without compromising their privacy. In the FEEL framework, edge devices periodically transmit high-dimensional stochastic gradients to the edge server, where these gradients are aggregated and used to update a global model. When the edge devices share the same communication medium, the multiple access channel (MAC) from the devices to the edge server induces a communication bottleneck. To overcome this bottleneck, an efficient broadband analog transmission scheme has been recently proposed, featuring the aggregation of analog modulated gradients (or local models) via the waveform-superposition property of the wireless medium. However, the assumed linear analog modulation makes it difficult to deploy this technique in modern wireless systems that exclusively use digital modulation. To address this issue, we propose in this work a novel digital version of broadband over-the-air aggregation, called one-bit broadband digital aggregation (OBDA). The new scheme features one-bit gradient quantization followed by digital quadrature amplitude modulation (QAM) at edge devices and over-the-air majority-voting based decoding at edge server. We provide a comprehensive analysis of the effects of wireless channel hostilities (channel noise, fading, and channel estimation errors) on the convergence rate of the proposed FEEL scheme. The analysis shows that the hostilities slow down the convergence of the learning process by introducing a scaling factor and a bias term into the gradient norm. However, we show that all the negative effects vanish as the number of participating devices grows, but at a different rate for each type of channel hostility.

Journal article

Giaconi G, Gunduz D, Poor HV, 2020, Privacy-cost trade-offs in smart electricity metering systems, IET Smart Grid, Vol: 3, Pages: 596-604, ISSN: 2515-2947

Trade-offs between privacy and cost are studied for a smart grid consumer, whose electricity consumption is monitoredin almost real time by the utility provider (UP) through smart meter (SM) readings. It is assumed that an electrical battery isavailable to the consumer, which can be utilized both to achieve privacy and to reduce the energy cost by demand shaping.Privacy is measured via the mean squared distance between the SM readings and a target load profile, while time-of-use (ToU)pricing is considered to compute the cost incurred. The consumer can also sell electricity back to the UP to further improve theprivacy-cost trade-off. Two privacy-preserving energy management policies (EMPs) are proposed, which differ in the way the targetload profile is characterized. A more practical EMP, which optimizes the energy management less frequently, is also considered.Numerical results are presented to compare the privacy-cost trade-off of these EMPs, considering various privacy indicators.

Journal article

Gunduz D, Oechtering T, Hug G, Kundur D, Arani M, Teng Fet al., 2020, Guest editorial: privacy and security in smart grids, IET Smart Grid, Vol: 3, Pages: 549-550, ISSN: 2515-2947

Journal article

Hameed MZ, Gyorgy A, Gunduz D, 2020, The best defense is a good offense: adversarial attacks to avoid modulation detection, IEEE Transactions on Information Forensics and Security, Vol: 16, Pages: 1074-1087, ISSN: 1556-6013

We consider a communication scenario, in which an intruder tries to determine the modulation scheme of the intercepted signal. Our aim is to minimize the accuracy of the intruder, while guaranteeing that the intended receiver can still recover the underlying message with the highest reliability. This is achieved by perturbing channel input symbols at the encoder,similarly to adversarial attacks against classifiers in machine learning. In image classification, the perturbation is limited to be imperceptible to a human observer, while in our case the perturbation is constrained so that the message can still be reliably decoded by the legitimate receiver, which is oblivious to the perturbation. Simulation results demonstrate the viability of our approach to make wireless communication secure against state-of-the-art intruders (using deep learning or decision trees)with minimal sacrifice in the communication performance. On he other hand, we also demonstrate that using diverse training data and curriculum learning can significantly boost the accuracy of the intruder.

Journal article

Temesgene DA, Miozzo M, Gunduz D, Dini Pet al., 2020, Distributed deep reinforcement learning for functional split control in energy harvesting virtualized small cells, IEEE Transactions on Sustainable Computing, ISSN: 2377-3782

To meet the growing quest for enhanced network capacity, mobile network operators (MNOs) are deploying dense infrastructures of small cells. This, in turn, increases the power consumption of mobile networks, thus impacting the environment. As a result, we have seen a recent trend of powering mobile networks with harvested ambient energy to achieve both environmental and cost benefits. In this paper, we consider a network of virtualized small cells (vSCs) powered by energy harvesters and equipped with rechargeable batteries, which can opportunistically offload baseband (BB) functions to a grid-connected edge server depending on their energy availability. We formulate the corresponding grid energy and traffic drop rate minimization problem, and propose a distributed deep reinforcement learning (DDRL) solution. Coordination among vSCs is enabled via the exchange of battery state information. The evaluation of the network performance in terms of grid energy consumption and traffic drop rate confirms that enabling coordination among the vSCs via knowledge exchange achieves a performance close to the optimal. Numerical results also confirm that the proposed DDRL solution provides higher network performance, better adaptation to the changing environment, and higher cost savings with respect to a tabular multi-agent reinforcement learning (MRL) solution used as a benchmark.

Journal article

Zhao J, Mohammadi Amiri M, Gunduz D, 2020, Multi-antenna coded content delivery with caching: a low-complexity solution, IEEE Transactions on Wireless Communications, Vol: 19, Pages: 7484-7497, ISSN: 1536-1276

We study downlink beamforming in a single-cellnetwork with a multi-antenna base station serving cache-enabledusers. Assuming a library of files with a common rate, we formulate the minimum transmit power with proactive caching andcoded delivery as a non-convex optimization problem. While thismultiple multicast problem can be efficiently solved by successiveconvex approximation (SCA), the complexity of the problemgrows exponentially with the number of subfiles delivered toeach user in each time slot, which itself grows exponentially withthe number of users. We introduce a low-complexity alternativethrough time-sharing that limits the number of subfiles receivedby a user in each time slot. We then consider the joint design ofbeamforming and content delivery with sparsity constraints tolimit the number of subfiles received by a user in each time slot.Numerical simulations show that the low-complexity scheme hasonly a small performance gap to that obtained by solving the jointproblem with sparsity constraints, and outperforms state-of-theart results at all signal-to-noise ratio (SNR) and rate values witha sufficient number of transmit antennas. A lower bound onthe achievable degrees-of-freedom (DoF) of the low-complexityscheme is derived to characterize its performance in the highSNR regime.

Journal article

Mital N, Gunduz D, Ling C, 2020, Coded caching in a multi-server system with random topology, IEEE Transactions on Communications, Vol: 68, Pages: 4620-4631, ISSN: 0090-6778

Cache-aided content delivery is studied in a multi-server system with P servers and K users, each equipped with a local cache memory. In the delivery phase, each user connects randomly to any ρ out of P servers. Thanks to the availability of multiple servers, which model small-cell base stations (SBSs), demands can be satisfied with reduced storage capacity at each server and reduced delivery rate per server; however, this also leads to reduced multicasting opportunities compared to the single-server scenario. A joint storage and proactive caching scheme is proposed, which exploits coded storage across the servers, uncoded cache placement at the users, and coded delivery. The delivery latency is studied for both successive and parallel transmissions from the servers. It is shown that, with successive transmissions the achievable average delivery latency is comparable to the one achieved in the single-server scenario, while the gap between the two depends on ρ, the available redundancy across the servers, and can be reduced by increasing the storage capacity at the SBSs. The optimality of the proposed scheme with uncoded cache placement and MDS-coded server storage is also proved for successive transmissions.

Journal article

Erdemir E, Dragotti PL, Gunduz D, 2020, Privacy-aware time-series data sharing with deep reinforcement learning, IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURIT, Vol: 16, Pages: 389-401, ISSN: 1556-6013

Internet of things (IoT) devices are becoming increasingly popular thanks to many new services and applications they offer. However, in addition to their many benefits, they raise privacy concerns since they share fine-grained time-series user data with untrusted third parties. In this work, we study the privacy-utility trade-off (PUT) in time-series data sharing. Existing approaches to PUT mainly focus on a single data point; however, temporal correlations in time-series data introduce new challenges. Methods that preserve the privacy for the current time may leak significant amount of information at the trace level as the adversary can exploit temporal correlations in a trace. We consider sharing the distorted version of a user’s true data sequence with an untrusted third party. We measure the privacy leakage by the mutual information between the user’s true data sequence and shared version. We consider both the instantaneous and average distortion between the two sequences, under a given distortion measure, as the utility loss metric. To tackle the history-dependent mutual information minimization, we reformulate the problem as a Markov decision process (MDP), and solve it using asynchronous actor-critic deep reinforcement learning (RL). We evaluate the performance of the proposed solution in location trace privacy on both synthetic and GeoLife GPS trajectory datasets. For the latter, we show the validity of our solution by testing the privacy of the released location trajectory against an adversary network.

Journal article

Faqir OJ, Kerrigan EC, Gunduz D, 2020, Information transmission bounds between moving terminals, IEEE Communications Letters, Vol: 24, Pages: 1410-1413, ISSN: 1089-7798

In networks of mobile autonomous agents, e.g. for data acquisition, we may wish to maximize data transfer or to reliably transfer a minimum amount of data, subject to quality of service or energy constraints. These requirements can be guaranteed through both offline node design/specifications and online trajectory/communications design. Regardless of the distance between them, for a stationary point-to-point transmitter-receiver pair communicating across a single link under average power constraints, the total data transfer is unbounded as time tends to infinity. In contrast, we show that if the transmitter/receiver is moving at any constant speed away from each other, then the maximum transmittable data is bounded. Although general closed-form expressions as a function of communication and mobility profile parameters do not yet exist, we provide closed-form expressions for particular cases, such as ideal free space path loss. Under more general scenarios we instead give lower bounds on the total transmittable information across a single link between mobile nodes.

Journal article

Sreekumar S, Cohen A, Gündüz D, 2020, Privacy-Aware Distributed Hypothesis Testing, Entropy, Vol: 22, Pages: 665-665

<jats:p>A distributed binary hypothesis testing (HT) problem involving two parties, a remote observer and a detector, is studied. The remote observer has access to a discrete memoryless source, and communicates its observations to the detector via a rate-limited noiseless channel. The detector observes another discrete memoryless source, and performs a binary hypothesis test on the joint distribution of its own observations with those of the observer. While the goal of the observer is to maximize the type II error exponent of the test for a given type I error probability constraint, it also wants to keep a private part of its observations as oblivious to the detector as possible. Considering both equivocation and average distortion under a causal disclosure assumption as possible measures of privacy, the trade-off between the communication rate from the observer to the detector, the type II error exponent, and privacy is studied. For the general HT problem, we establish single-letter inner bounds on both the rate-error exponent-equivocation and rate-error exponent-distortion trade-offs. Subsequently, single-letter characterizations for both trade-offs are obtained (i) for testing against conditional independence of the observer’s observations from those of the detector, given some additional side information at the detector; and (ii) when the communication rate constraint over the channel is zero. Finally, we show by providing a counter-example where the strong converse which holds for distributed HT without a privacy constraint does not hold when a privacy constraint is imposed. This implies that in general, the rate-error exponent-equivocation and rate-error exponent-distortion trade-offs are not independent of the type I error probability constraint.</jats:p>

Journal article

Zhao J, Gunduz D, Simeone O, Gomez-Barquero Det al., 2020, Non-orthogonal unicast and broadcast transmission via joint beamforming and LDM in cellular networks, IEEE Transactions on Broadcasting, Vol: 66, Pages: 216-228, ISSN: 0018-9316

Limited bandwidth resources and higher energy efficiency requirements motivate incorporating multicast and broadcast transmission into the next-generation cellular network architectures, particularly for multimedia streaming applications. Layered division multiplexing (LDM), a form of non-orthogonal multiple access (NOMA), can potentially improve unicast throughput and broadcast coverage with respect to traditional orthogonal frequency division multiplexing (FDM) or time division multiplexing (TDM), by simultaneously using the same frequency and time resources for multiple unicast or broadcast transmissions. In this paper, the performance of LDM-based unicast and broadcast transmission in a cellular network is studied by assuming a single frequency network (SFN) operation for the broadcast layer, while allowing arbitrarily clustered cooperation among the base stations (BSs) for the transmission of unicast data streams. Beamforming and power allocation between unicast and broadcast layers, the so-called injection level in the LDM literature, are optimized with the aim of minimizing the sum-power under constraints on the user-specific unicast rates and on the common broadcast rate. The effects of imperfect channel coding and imperfect channel state information (CSI) are also studied to gain insights into robust implementation in practical systems. The non-convex optimization problem is tackled by means of successive convex approximation (SCA) techniques. Performance upper bounds are also presented by means of the S-procedure followed by semidefinite relaxation (SDR). Finally, a dual decomposition-based solution is proposed to facilitate an efficient distributed implementation of LDM in each of the SCA subproblems, where the unicast beamforming vectors can be obtained locally by the cooperating BSs. Numerical results are presented, which show the tightness of the proposed bounds and hence the near-optimality of the proposed solutions.

Journal article

Amiri MM, Gunduz D, Kulkarni SR, Poor HVet al., 2020, Update Aware Device Scheduling for Federated Learning at the Wireless Edge, Pages: 2598-2603, ISSN: 2157-8095

We study federated learning (FL) at the wireless edge, where power-limited devices with local datasets train a joint model with the help of a remote parameter server (PS). We assume that the devices are connected to the PS through a bandwidth-limited shared wireless channel. At each iteration of FL, a subset of the devices are scheduled to transmit their local model updates to the PS over orthogonal channel resources. We design novel scheduling policies, that decide on the subset of devices to transmit at each round not only based on their channel conditions, but also on the significance of their local model updates. Numerical results show that the proposed scheduling policy provides a better long-term performance than scheduling policies based only on either of the two metrics individually. We also observe that when the data is independent and identically distributed (i.i.d.) across devices, selecting a single device at each round provides the best performance, while when the data distribution is non-i.i.d., more devices should be scheduled.

Conference paper

Sreekumar S, Gunduz D, 2020, Strong Converse for Testing Against Independence over a Noisy channel, Pages: 1283-1288, ISSN: 2157-8095

A distributed binary hypothesis testing (HT) problem over a noisy (discrete and memoryless) channel studied previously by the authors is investigated from the perspective of the strong converse property. It was shown by Ahlswede and Csiszar that a strong converse holds in the above setting when the channel is rate-limited and noiseless. Motivated by this observation, we show that the strong converse continues to hold in the noisy channel setting for a special case of HT known as testing against independence (TAI), under the assumption that the channel transition matrix has non-zero elements. The proof utilizes the blowing up lemma and the recent change of measure technique of Tyagi and Watanabe as the key tools.

Conference paper

Hasircioglu B, Gomez-Vilardebo J, Gunduz D, 2020, Bivariate Polynomial Coding for Straggler Exploitation with Heterogeneous Workers, Pages: 251-256, ISSN: 2157-8095

Polynomial coding has been proposed as a solution to the straggler mitigation problem in distributed matrix multiplication. Previous works employ univariate polynomials to encode matrix partitions. Such schemes greatly improve the speed of distributed computing systems by making the task completion time to depend only on the fastest workers. However, they completely ignore the work done by the slowest workers resulting in inefficient use of computing resources. In order to exploit the partial computations of the slower workers, we further decompose the overall matrix multiplication task into even smaller subtasks, and we propose bivariate polynomial codes. We show that these codes are a more natural choice to accommodate the additional decomposition of subtasks, and to exploit the heterogeneous storage and computation resources at workers. However, in contrast to univariate polynomial decoding, guarantying decodability with multivariate interpolation is much harder. We propose two bivariate polynomial coding schemes and study their decodability conditions. Our numerical results show that bivariate polynomial coding considerably reduces the computation time of distributed matrix multiplication.

Conference paper

Ozfatura M, Gunduz D, 2020, Mobility-aware coded storage and delivery, IEEE Transactions on Communications, Vol: 68, Pages: 3275-3285, ISSN: 0090-6778

We consider a cache-enabled heterogeneous cellular network, where mobile users (MUs) connect to multiple cache-enabled small-cell base stations (SBSs) during a video downloading session. SBSs can deliver these requests using their local cache contents as well as by downloading them from a macro-cell base station (MBS), which has access to the file library. We introduce a novel mobility-aware content storage and delivery scheme, which jointly exploits coded storage at the SBSs and coded delivery from the MBS to reduce the backhaul load from the MBS to the SBSs. We show that the proposed scheme provides a significant reduction both in the backhaul load when the cache capacity is sufficiently large, and in the number of sub-files required. Overall, for practical scenarios, in which the number of sub-files that can be created is limited either by the size of the files, or by the protocol overhead, the proposed coded caching and delivery scheme decidedly outperforms state-of-the-art alternatives. Finally, we show that the benefits of the proposed scheme also extends to scenarios with non-uniform file popularities and arbitrary mobility patterns.

Journal article

Ozfatura E, Ulukus S, Gündüz D, 2020, Straggler-aware distributed learning: communication–computation latency trade-off, Entropy: international and interdisciplinary journal of entropy and information studies, Vol: 22, Pages: 544-544, ISSN: 1099-4300

When gradient descent (GD) is scaled to many parallel workers for large-scale machine learning applications, its per-iteration computation time is limited by straggling workers. Straggling workers can be tolerated by assigning redundant computations and/or coding across data and computations, but in most existing schemes, each non-straggling worker transmits one message per iteration to the parameter server (PS) after completing all its computations. Imposing such a limitation results in two drawbacks: over-computation due to inaccurate prediction of the straggling behavior, and under-utilization due to discarding partial computations carried out by stragglers. To overcome these drawbacks, we consider multi-message communication (MMC) by allowing multiple computations to be conveyed from each worker per iteration, and propose novel straggler avoidance techniques for both coded computation and coded communication with MMC. We analyze how the proposed designs can be employed efficiently to seek a balance between the computation and communication latency. Furthermore, we identify the advantages and disadvantages of these designs in different settings through extensive simulations, both model-based and real implementation on Amazon EC2 servers, and demonstrate that proposed schemes with MMC can help improve upon existing straggler avoidance schemes.

Journal article

Popovski P, Simeone O, Boccardi F, Gunduz D, Sahin Oet al., 2020, Semantic-effectiveness filtering and control for post-5G wireless connectivity, Journal of the Indian Institute of Science, Vol: 100, Pages: 435-443, ISSN: 0970-4140

The traditional role of a communication engineer is to address the technical problem of transporting bits reliably over a noisy channel. With the emergence of 5G, and the availability of a variety of competing and coexisting wireless systems, wireless connectivity is becoming a commodity. This article argues that communication engineers in the post-5G era should extend the scope of their activity in terms of design objectives and constraints beyond connectivity to encompass the semantics of the transferred bits within the given applications and use cases. To provide a platform for semantic-aware connectivity solutions, this paper introduces the concept of a semantic-effectiveness (SE) plane as a core part of future communication architectures. The SE plane augments the protocol stack by providing standardized interfaces that enable information filtering and direct control of functionalities at all layers of the protocol stack. The advantages of the SE plane are described in the perspective of recent developments in 5G, and illustrated through a number of example applications. The introduction of a SE plane may help replacing the current “next-G paradigm” in wireless evolution with a framework based on continuous improvements and extensions of the systems and standards.

Journal article

Boloursaz Mashhadi M, Gunduz D, 2020, Deep learning for massive MIMO channel state acquisition and feedback, Journal of the Indian Institute of Science, Vol: 100, Pages: 369-382, ISSN: 0970-4140

Massive multiple-input multiple-output (MIMO) systems are a main enabler of the excessive throughput requirements in 5G and future generation wireless networks as they can serve many users simultaneously with high spectral and energy efficiency. To achieve this massive MIMO systems require accurate and timely channel state information (CSI), which is acquired by a training process that involves pilot transmission, CSI estimation, and feedback. This training process incurs a training overhead, which scales with the number of antennas, users, and subcarriers. Reducing the training overhead in massive MIMO systems has been a major topic of research since the emergence of the concept. Recently, deep learning (DL)-based approaches have been proposed and shown to provide significant reduction in the CSI acquisition and feedback overhead in massive MIMO systems compared to traditional techniques. In this paper, we present an overview of the state-of-the-art DL architectures and algorithms used for CSI acquisition and feedback, and provide further research directions.

Journal article

Kurka DB, Gunduz D, 2020, DeepJSCC-f: deep joint source-channel coding of images with feedback, IEEE Journal on Selected Areas in Information Theory, Vol: 1, Pages: 178-193, ISSN: 2641-8770

We consider wireless transmission of images in the presence of channel output feedback. From a Shannon theoretic perspective feedback does not improve the asymptotic end-to-end performance, and separate source coding followed by capacity-achieving channel coding, which ignores the feedback signal, achieves the optimal performance. It is well known that separation is not optimal in the practical finite blocklength regime; however, there are no known practical joint source-channel coding (JSCC) schemes that can exploit the feedback signal and surpass the performance of separation-based schemes. Inspired by the recent success of deep learning methods for JSCC, we investigate how noiseless or noisy channel output feedback can be incorporated into the transmission system to improve the reconstruction quality at the receiver. We introduce an autoencoder-based JSCC scheme, which we call DeepJSCC-f, that exploits the channel output feedback, and provides considerable improvements in terms of the end-to-end reconstruction quality for fixed-length transmission, or in terms of the average delay for variable-length transmission. To the best of our knowledge, this is the first practical JSCC scheme that can fully exploit channel output feedback, demonstrating yet another setting in which modern machine learning techniques can enable the design of new and efficient communication methods that surpass the performance of traditional structured coding-based designs.

Journal article

Jankowski M, Gunduz D, Mikolajczyk K, 2020, Joint Device-Edge Inference over Wireless Links with Pruning

We propose a joint feature compression and transmission scheme for efficient inference at the wireless network edge. Our goal is to enable efficient and reliable inference at the edge server assuming limited computational resources at the edge device. Previous work focused mainly on feature compression, ignoring the computational cost of channel coding. We incorporate the recently proposed deep joint source-channel coding (DeepJSCC) scheme, and combine it with novel filter pruning strategies aimed at reducing the redundant complexity from neural networks. We evaluate our approach on a classification task, and show improved results in both end-To-end reliability and workload reduction at the edge device. This is the first work that combines DeepJSCC with network pruning, and applies it to image classification over the wireless edge.

Conference paper

Mohammadi Amiri M, Gunduz D, 2020, Federated learning over wireless fading channels, IEEE Transactions on Wireless Communications, Vol: 19, Pages: 3546-3557, ISSN: 1536-1276

We study federated machine learning at the wirelessnetwork edge, where limited power wireless devices, each withits own dataset, build a joint model with the help of a remoteparameter server (PS). We consider a bandwidth-limited fadingmultiple access channel (MAC) from the wireless devices to thePS, and propose various techniques to implement distributedstochastic gradient descent (DSGD) over this shared noisywireless channel. We first propose a digital DSGD (D-DSGD)scheme, in which one device is selected opportunistically fortransmission at each iteration based on the channel conditions;the scheduled device quantizes its gradient estimate to a finitenumber of bits imposed by the channel condition, and transmitsthese bits to the PS in a reliable manner. Next, motivated bythe additive nature of the wireless MAC, we propose a novelanalog communication scheme, referred to as thecompressedanalogDSGD (CA-DSGD), where the devices first sparsifytheir gradient estimates while accumulating error from previousiterations, and project the resultant sparse vector into a low-dimensional vector for bandwidth reduction. We also design apower allocation scheme to align the received gradient vectorsat the PS in an efficient manner. Numerical results show thatD-DSGD outperforms other digital approaches in the literature;however, in general the proposed CA-DSGD algorithm convergesfaster than the D-DSGD scheme, and reaches a higher level ofaccuracy. We have observed that the gap between the analogand digital schemes increases when the datasets of devices arenot independent and identically distributed (i.i.d.). Furthermore,the performance of the CA-DSGD scheme is shown to be robustagainst imperfect channel state information (CSI) at the devices.Overall these results show clear advantages for

Journal article

Amiri MM, Gunduz D, 2020, Machine learning at the wireless edge: distributed stochastic gradient descent over-the-air, IEEE Transactions on Signal Processing, Vol: 68, Pages: 2155-2169, ISSN: 1053-587X

We study federated machine learning (ML) at the wireless edge, where power- and bandwidth-limited wireless devices with local datasets carry out distributed stochastic gradient descent (DSGD) with the help of a parameter server (PS). Standard approaches assume separate computation and communication, where local gradient estimates are compressed and transmitted to the PS over orthogonal links. Following this digital approach, we introduce D-DSGD, in which the wireless devices employ gradient quantization and error accumulation, and transmit their gradient estimates to the PS over a multiple access channel (MAC). We then introduce a novel analog scheme, called A-DSGD, which exploits the additive nature of the wireless MAC for over-the-air gradient computation, and provide convergence analysis for this approach. In A-DSGD, the devices first sparsify their gradient estimates, and then project them to a lower dimensional space imposed by the available channel bandwidth. These projections are sent directly over the MAC without employing any digital code. Numerical results show that A-DSGD converges faster than D-DSGD thanks to its more efficient use of the limited bandwidth and the natural alignment of the gradient estimates over the channel. The improvement is particularly compelling at low power and low bandwidth regimes. We also illustrate for a classification problem that, A-DSGD is more robust to bias in data distribution across devices, while D-DSGD significantly outperforms other digital schemes in the literature. We also observe that both D-DSGD and A-DSGD perform better with the number of devices, showing their ability in harnessing the computation power of edge devices.

Journal article

This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.

Request URL: http://wlsprd.imperial.ac.uk:80/respub/WEB-INF/jsp/search-html.jsp Request URI: /respub/WEB-INF/jsp/search-html.jsp Query String: respub-action=search.html&id=00761710&limit=30&person=true