Publications
374 results found
Abad MSH, Ozfatura E, Gunduz D, et al., 2020, HIERARCHICAL FEDERATED LEARNING ACROSS HETEROGENEOUS CELLULAR NETWORKS, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Publisher: IEEE, Pages: 8866-8870, ISSN: 1520-6149
- Author Web Link
- Cite
- Citations: 135
Amiri MM, Gunduz D, Kulkarni SR, et al., 2020, Update Aware Device Scheduling for Federated Learning at the Wireless Edge, IEEE International Symposium on Information Theory (ISIT), Publisher: IEEE, Pages: 2598-2603
- Author Web Link
- Cite
- Citations: 34
Sreekumar S, Gunduz D, 2020, Strong Converse for Testing Against Independence over a Noisy channel, IEEE International Symposium on Information Theory (ISIT), Publisher: IEEE, Pages: 1283-1288
- Author Web Link
- Cite
- Citations: 5
Hasircioglu B, Gomez-Vilardebo J, Gunduz D, 2020, Bivariate Polynomial Coding for Straggler Exploitation with Heterogeneous Workers, IEEE International Symposium on Information Theory (ISIT), Publisher: IEEE, Pages: 251-256
- Author Web Link
- Cite
- Citations: 5
Sun Y, Zhou S, Gunduz D, 2020, Energy-Aware Analog Aggregation for Federated Learning with Redundant Data, IEEE International Conference on Communications (IEEE ICC) / Workshop on NOMA for 5G and Beyond, Publisher: IEEE, ISSN: 1550-3607
- Author Web Link
- Cite
- Citations: 38
Mashhadi MB, Yang Q, Gunduz D, 2020, CNN-BASED ANALOG CSI FEEDBACK IN FDD MIMO-OFDM SYSTEMS, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Publisher: IEEE, Pages: 8579-8583, ISSN: 1520-6149
- Author Web Link
- Cite
- Citations: 12
Shao Y, Gunduz D, Liew SC, 2020, Federated Edge Learning with Misaligned Over-The-Air Computation, 22nd IEEE International Workshop on Signal Processing Advances in Wireless Communications (IEEE SPAWC), Publisher: IEEE, Pages: 236-240, ISSN: 2325-3789
Jankowski M, Gunduz D, Mikolajczyk K, 2020, DEEP JOINT SOURCE-CHANNEL CODING FOR WIRELESS IMAGE RETRIEVAL, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Publisher: IEEE, Pages: 5070-5074, ISSN: 1520-6149
- Author Web Link
- Cite
- Citations: 17
Marchioro T, Laurenti N, Gunduz D, 2020, ADVERSARIAL NETWORKS FOR SECURE WIRELESS COMMUNICATIONS, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Publisher: IEEE, Pages: 8748-8752, ISSN: 1520-6149
- Author Web Link
- Cite
- Citations: 6
Jankowski M, Gunduz D, Mikolajczyk K, 2020, Joint Device-Edge Inference over Wireless Links with Pruning, PROCEEDINGS OF THE 21ST IEEE INTERNATIONAL WORKSHOP ON SIGNAL PROCESSING ADVANCES IN WIRELESS COMMUNICATIONS (IEEE SPAWC2020), ISSN: 2325-3789
- Author Web Link
- Cite
- Citations: 15
Kurka DB, Gunduz D, 2020, DEEP JOINT SOURCE-CHANNEL CODING OF IMAGES WITH FEEDBACK, IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Publisher: IEEE, Pages: 5235-5239, ISSN: 1520-6149
- Author Web Link
- Cite
- Citations: 8
Amiri MM, Gunduz D, 2019, Computation scheduling for distributed machine learning with straggling workers, IEEE Transactions on Signal Processing, Vol: 67, Pages: 6270-6284, ISSN: 1053-587X
We study scheduling of computation tasks across n workers in a large scale distributed learning problem with the help of a master. Computation and communication delays are assumed to be random, and redundant computations are assigned to workers in order to tolerate stragglers. We consider sequential computation of tasks assigned to a worker, while the result of each computation is sent to the master right after its completion. Each computation round, which can model an iteration of the stochastic gradient descent (SGD) algorithm, is completed once the master receives k distinct computations, referred to as the computation target. Our goal is to characterize the average completion time as a function of the computation load, which denotes the portion of the dataset available at each worker, and the computation target. We propose two computation scheduling schemes that specify the tasks assigned to each worker, as well as their computation schedule, i.e., the order of execution. Assuming a general statistical model for computation and communication delays, we derive the average completion time of the proposed schemes. We also establish a lower bound on the minimum average completion time by assuming prior knowledge of the random delays. Experimental results carried out on Amazon EC2 cluster show a significant reduction in the average completion time over existing coded and uncoded computing schemes. It is also shown numerically that the gap between the proposed scheme and the lower bound is relatively small, confirming the efficiency of the proposed scheduling design.
Sreekumar S, Gunduz D, 2019, Distributed hypothesis testing over discrete memoryless channels, IEEE Transactions on Information Theory, Vol: 66, Pages: 2044-2066, ISSN: 0018-9448
A distributed binary hypothesis testing (HT) problem involving two parties, one referred to as the observer and the other as the detector is studied. The observer observes a discrete memoryless source (DMS) and communicates its observations to the detector over a discrete memoryless channel (DMC). The detector observes another DMS correlated with that at the observer, and performs a binary HT on the joint distribution of the two DMS’s using its own observed data and the information received from the observer. The trade-off between the type I error probability and the type II error-exponent of the HT is explored. Single-letter lower bounds on the optimal type II errorexponent are obtained by using two different coding schemes, a separate HT and channel coding scheme and a joint HT and channel coding scheme based on hybrid coding for the matched bandwidth case. Exact single-letter characterization of the same is established for the special case of testing against conditional independence, and it is shown to be achieved by the separate HT and channel coding scheme. An example is provided where the joint scheme achieves a strictly better performance than the separation based scheme.
Gesbert D, Gunduz D, de Kerret P, et al., 2019, Guest Editorial Special Issue on Machine Learning in Wireless Communication-Part 2, IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, Vol: 37, Pages: 2409-2412, ISSN: 0733-8716
- Author Web Link
- Cite
- Citations: 1
Yang Q, Mohammadi Amiri M, Gunduz D, 2019, Audience-retention-rate-aware caching and coded video delivery with asynchronous demands, IEEE Transactions on Communications, Vol: 67, Pages: 7088-7102, ISSN: 0090-6778
Most of the current literature on coded cachingfocus on a static scenario, in which a fixed number of userssynchronously place their requests from a content library, andthe performance is measured in terms of the latency in satisfyingall of these requests. In practice, however, users start watching anonline video content asynchronously over time, and often abortwatching a video before it is completed. The latter behaviour iscaptured by the notion of audience retention rate, which measuresthe portion of a video content watched on average. In order tobring coded caching one step closer to practice, asynchronoususer demands are considered in this paper, by allowing userdemands to arrive randomly over time, and both the popularityof video files, and the audience retention rates are taken intoaccount. A decentralized partial coded delivery (PCD) schemeis proposed, and two cache allocation schemes are employed;namely homogeneous cache allocation (HoCA) and heterogeneouscache allocation (HeCA), which allocate users’ caches amongdifferent chunks of the video files in the library. Numerical resultsvalidate that the proposed PCD scheme, either with HoCA orHeCA, outperforms conventional uncoded caching as well as thestate-of-the-art decentralized caching schemes, which consideronly the file popularities, and are designed for synchronousdemand arrivals. An information-theoretical lower bound on theaverage delivery rate is also presented.
Gunduz D, de Kerret P, Sidiropoulos ND, et al., 2019, Machine learning in the air, IEEE Journal on Selected Areas in Communications, Vol: 37, Pages: 2184-2199, ISSN: 0733-8716
Thanks to the recent advances in processing speed, data acquisition and storage, machine learning (ML) is penetrating every facet of our lives, and transforming research in many areas in a fundamental manner. Wireless communications is another success story - ubiquitous in our lives, from handheld devices to wearables, smart homes, and automobiles. While recent years have seen a flurry of research activity in exploiting ML tools for various wireless communication problems, the impact of these techniques in practical communication systems and standards is yet to be seen. In this paper, we review some of the major promises and challenges of ML in wireless communication systems, focusing mainly on the physical layer. We present some of the most striking recent accomplishments that ML techniques have achieved with respect to classical approaches, and point to promising research directions where ML is likely to make the biggest impact in the near future. We also highlight the complementary problem of designing physical layer techniques to enable distributed ML at the wireless network edge, which further emphasizes the need to understand and connect ML with fundamental concepts in wireless communications.
Yang Q, Mashhadi MB, Gunduz D, 2019, Deep Convolutional Compression For Massive MIMO CSI Feedback, 2019 IEEE 29th International Workshop on Machine Learning for Signal Processing (MLSP), Publisher: IEEE
Gesbert D, Gunduz D, de Kerret P, et al., 2019, Special Issue on Machine Learning in Wireless Communication-Part I, IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, Vol: 37, Pages: 2181-2183, ISSN: 0733-8716
- Author Web Link
- Cite
- Citations: 3
Bourtsoulatze E, Kurka DB, Gunduz D, 2019, Deep joint source-channel coding for wireless image transmission, IEEE Transactions on Cognitive Communications and Networking, Vol: 5, Pages: 567-579, ISSN: 2332-7731
We propose a joint source and channel coding (JSCC) technique for wireless image transmission that does not rely on explicit codes for either compression or error correction; instead, it directly maps the image pixel values to the complex-valued channel input symbols. We parameterize the encoder and decoder functions by two convolutional neural networks (CNNs), which are trained jointly, and can be considered as an autoencoder with a non-trainable layer in the middle that represents the noisy communication channel. Our results show that the proposed deep JSCC scheme outperforms digital transmission concatenating JPEG or JPEG2000 compression with a capacity achieving channel code at low signal-to-noise ratio (SNR) and channel bandwidth values in the presence of additive white Gaussian noise (AWGN). More strikingly, deep JSCC does not suffer from the “cliff effect”, and it provides a graceful performance degradation as the channel SNR varies with respect to the SNR value assumed during training. In the case of a slow Rayleigh fading channel, deep JSCC learns noise resilient coded representations and significantly outperforms separation-based digital communication at all SNR and channel bandwidth values.
Cao D, Zhang D, Chen P, et al., 2019, Coded caching with asymmetric cache sizes and link qualities: The two-user case, IEEE Transactions on Communications, Vol: 67, Pages: 6112-6126, ISSN: 0090-6778
Centralized coded caching problem is studied for the two-user scenario, considering heterogeneous cache capacities at the users and private channels from the server to the users, in addition to a shared channel. Optimal caching and delivery strategies that minimize the worst-case delivery latency are presented for an arbitrary number of files. The converse proof follows from the sufficiency of file-index-symmetric caching and delivery codes, while the achievability is obtained through memory-sharing among a number of special memory–capacity pairs. The optimal scheme is shown to exploit the private link capacities by transmitting part of the corresponding user‘s request in an uncoded fashion. When there are no private links, the results presented here improve upon the two known results in the literature, namely, i) equal cache capacities and arbitrary number of files; and ii) unequal cache capacities and two files. The results are then extended to the caching problem with heterogeneous distortion requirements.
Mital N, Kralevska K, Ling C, et al., 2019, Practical Functional Regenerating Codes for Broadcast Repair of Multiple Nodes, IEEE International Symposium on Information Theory (ISIT), Publisher: IEEE, Pages: 221-225
Ozfatura E, Gunduz D, Ulukus S, 2019, Speeding Up Distributed Gradient Descent by Utilizing Non-persistent Stragglers, IEEE International Symposium on Information Theory (ISIT), Publisher: IEEE, Pages: 2729-2733
Li Z, Oechtering TJ, Gunduz D, 2019, Privacy against a hypothesis testing adversary, IEEE Transactions on Information Forensics and Security, Vol: 14, Pages: 1567-1581, ISSN: 1556-6013
Privacy against an adversary (AD) that tries to detect the underlying privacy-sensitive data distribution is studied. The original data sequence is assumed to come from one of the two known distributions, and the privacy leakage is measured by the probability of error of the binary hypothesis test carried out by the AD. A management unit (MU) is allowed to manipulate the original data sequence in an online fashion, while satisfying an average distortion constraint. The goal of the MU is to maximize the minimal type II probability of error subject to a constraint on the type I probability of error assuming an adversarial Neyman-Pearson test, or to maximize the minimal error probability assuming an adversarial Bayesian test. The asymptotic exponents of the maximum minimal type II probability of error and the maximum minimal error probability are shown to be characterized by a Kullback-Leibler divergence rate and a Chernoff information rate, respectively. Privacy performances of particular management policies, the memoryless hypothesis-aware policy and the hypothesis-unaware policy with memory, are compared. The proposed formulation can also model adversarial example generation with minimal data manipulation to fool classifiers. Lastly, the results are applied to a smart meter privacy problem, where the user’s energy consumption is manipulated by adaptively using a renewable energy source in order to hide user’s activity from the energy provider.
Mohammadi Amiri M, Gunduz D, 2019, Computation scheduling for distributed machine learning with straggling workers, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing, Publisher: Institute of Electrical and Electronics Engineers, Pages: 8177-8181, ISSN: 2379-190X
We study scheduling of computation tasks acrossnworkers in a large scale distributed learning problem. Computa-tion speeds of the workers are assumed to be heterogeneous andunknown to the master, and redundant computations are assignedto the workers in order to tolerate straggling workers. We con-sider sequential computation and instantaneous communicationfrom each worker to the master, and each computation round,which can model a single iteration of the stochastic gradientdescent (SGD) algorithm, iscompletedonce the master receivesk≤ndistinct computations, referred to as thecomputationtarget. Our goal is to characterize theaverage completion timeas a function of thecomputation load, which denotes the portionof the dataset available at each worker, and the computationtarget. We propose two computation scheduling schemes thatspecify the computation tasks assigned to each worker, as wellas their order of execution. We also establish a lower bound onthe minimum average completion time. Numerical results showa significant reduction in the average computation time over theexisting coded and uncoded computing schemes.
Tao M, Gunduz D, Xu F, et al., 2019, Content caching and delivery in wireless radio access networks, IEEE Transactions on Communications, Vol: 67, Pages: 4724-4749, ISSN: 0090-6778
Today’s mobile data traffic is dominated by contentoriented traffic. Caching popular contents at the network edge can alleviate network congestion and reduce content delivery latency. This paper provides a comprehensive and unified study of caching and delivery techniques in wireless radio access networks (RANs) with caches at all edge nodes (ENs) and user equipments (UEs). Three cache-aided RAN architectures are considered: RANs without fronthaul, with dedicated fronthaul, and with wireless fronthaul. It first reviews in a tutorial nature how caching facilitates interference management in these networks by enabling interference cancellation (IC), zero-forcing (ZF), and interference alignment (IA). Then, two new delivery schemes are presented. One is for RANs with dedicated fronthaul, which considers centralized cache placement at the ENs but both centralized and decentralized placement at the UEs. This scheme combines IA, ZF, and IC together with soft-transfer fronthauling. The other is for RANs with wireless fronthaul, which considers decentralized cache placement at all nodes. It leverages the broadcast nature of wireless fronthaul to fetch not only uncached but also cached contents to boost transmission cooperation among the ENs. Numerical results show that both schemes outperform existing results for a wide range of system parameters, thanks to the various caching gains obtained opportunistically.
Rassouli B, Rosas FE, Gunduz D, 2019, Data disclosure under perfect sample privacy
Perfect data privacy seems to be in fundamental opposition to the economicaland scientific opportunities associated with extensive data exchange. Defyingthis intuition, this paper develops a framework that allows the disclosure ofcollective properties of datasets without compromising the privacy ofindividual data samples. We present an algorithm to build an optimal disclosurestrategy/mapping, and discuss it fundamental limits on finite andasymptotically large datasets. Furthermore, we present explicit expressions tothe asymptotic performance of this scheme in some scenarios, and study caseswhere our approach attains maximal efficiency. We finally discuss suboptimalschemes to provide sample privacy guarantees to large datasets with a reducedcomputational cost.
Rassouliy B, Gunduz D, 2019, Optimal utility-privacy trade-off with total variation distance as a privacy measure, IEEE Transactions on Information Forensics and Security, Vol: 15, Pages: 594-603, ISSN: 1556-6013
The total variation distance is proposed as a privacy measure in an information disclosure scenario when the goal is to reveal some information about available data in return of utility, while retaining the privacy of certain sensitive latent variables from the legitimate receiver. The total variation distance is introduced as a measure of privacy-leakage by showing that: i) it satis?es the post-processing and linkage inequalities, which makes it consistent with an intuitive notion of a privacy measure; ii) the optimal utility-privacy trade-off can be solved through a standard linear program when total variation distance is employed as the privacy measure; iii) it provides a bound on the privacy-leakage measured by mutual information, maximal leakage, or the improvement in an inference attack with a bounded cost function.
Ceran ET, Gunduz D, Gyorgy A, 2019, Average age of information with hybrid ARQ under a resource constraint, IEEE Transactions on Wireless Communications, Vol: 18, Pages: 1900-1913, ISSN: 1536-1276
Scheduling the transmission of status updates over an error-prone communication channel is studied in order to minimize the long-term average age of information at the destination under a constraint on the average number of transmissions at the source node. After each transmission, the source receives an instantaneous ACK/NACK feedback, and decides on the next update without prior knowledge on the success of future transmissions. The optimal scheduling policy is first studied under different feedback mechanisms when the channel statistics are known; in particular, the standard automatic repeat request (ARQ) and hybrid ARQ (HARQ) protocols are considered. The structural results are derived for the optimal policy under HARQ, while the optimal policy is determined analytically for ARQ. For the case of unknown environments, an average-cost reinforcement learning algorithm is proposed that learns the system parameters and the transmission policy in real time. The effectiveness of the proposed methods is verified through the numerical results.
Rassouli B, Rosas F, Gunduz D, 2019, Latent Feature Disclosure under Perfect Sample Privacy, 10th IEEE International Workshop on Information Forensics and Security (WIFS), Publisher: IEEE, ISSN: 2157-4766
Ceran ET, Gündüz D, György A, 2019, Reinforcement learning to minimize age of information with an energy Harvesting sensor with HARQ and sensing cost
The time average expected age of information (AoI) is studied for statusupdates sent from an energy-harvesting transmitter with a finite-capacitybattery. The optimal scheduling policy is first studied under differentfeedback mechanisms when the channel and energy harvesting statistics areknown. For the case of unknown environments, an average-cost reinforcementlearning algorithm is proposed that learns the system parameters and the statusupdate policy in real time. The effectiveness of the proposed methods isverified through numerical results.
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.