Mustafa Akdeniz

Dr. Mustafa Akdeniz


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 November 2020 in IEEE Journal on Selected Areas in Communications
Verified
Journal Article
Closed access
SPSaurav Prakash
Saurav Prakash
SDSagar Dhakal
Sagar Dhakal
MAMustafa Akdeniz
Mustafa Akdeniz
YYYair Yona
Yair Yona
STShilpa Talwar
Shilpa Talwar
NHNageen Himayat
Nageen Himayat
Federated learning enables training a global model from data located at the client nodes, without data sharing and moving client data to a centralized server. Performance of federated learning in a multi-access edge computing (MEC) network suffers from slow convergence due to heterogeneity and stochastic fluctuations in compute power and communication link qualities across clients. We propose a novel coded computing framework, CodedFedL, that injects structured coding redundancy into federated learning for mitigating stragglers and speeding up the training procedure. CodedFedL enables coded computing for non-linear federated learning by efficiently exploiting distributed kernel embedding via random Fourier features that transforms the training task into computationally favourable distributed linear regression. Furthermore, clients generate local parity datasets by coding over their local datasets, while the server combines them to obtain the global parity dataset. Gradient from the global parity dataset compensates for straggling gradients during training, and thereby speeds up convergence. For minimizing the epoch deadline time at the MEC server, we provide a tractable approach for finding the amount of coding redundancy and the number of local data points that a client processes during training, by exploiting the statistical properties of compute as well as communication delays. We also characterize the leakage in data privacy when clients share their local parity datasets with the server. Additionally, we analyze the convergence rate and iteration complexity of CodedFedL under simplifying assumptions, by treating CodedFedL as a stochastic gradient descent algorithm. Finally, for demonstrating gains that CodedFedL can achieve in practice, we conduct numerical experiments using practical network parameters and benchmark datasets, in which CodedFedL speeds up the overall training time by up to 15× in comparison to the benchmark schemes.

17 December 2018 in IEEE Transactions on Vehicular Technology
Verified
Journal Article
Closed access
MGMargarita Gapeyenko
Margarita Gapeyenko
VPVitaly Petrov
Vitaly Petrov
DMDmitri Moltchanov
Dmitri Moltchanov
MAMustafa Akdeniz
Mustafa Akdeniz
SASergey Andreev
Sergey Andreev
YKYevgeni Kourcheyavy
Yevgeni Kourcheyavy
Outage event caused by dynamic link blockage at millimeter-wave (mmWave) frequencies is a challenging problem for cell-edge users. To address it, 3GPP is currently working on multi-connectivity mechanisms that allow a user to remain connected to several mmWave access points simultaneously as well as switch between them in case its active connection drops. However, the actual number of such simultaneous links -- named the degree of multi-connectivity -- to reach the desired trade-off between the system design simplicity and the outage probability levels remains an open research question. In this work, we characterize the outage probability and spectral efficiency associated with different degrees of multi-connectivity in a typical 5G urban scenario, where the line-of-sight propagation path can be blocked by buildings as well as humans. These results demonstrate that the degrees of multi-connectivity of up to 4 offer higher relative gains. At the same time, performance improvements brought by the multi-connectivity degrees of 5 and higher are much lower. Our analytical framework can be further employed for the performance analysis of multi-connectivity-capable mmWave systems across their different deployment configurations.

23 May 2019 in IEEE Transactions on Mobile Computing
Verified
Journal Article
Closed access
MGMargarita Gapeyenko
Margarita Gapeyenko
ASAndrey Samuylov
Andrey Samuylov
MGMikhail Gerasimenko
Mikhail Gerasimenko
DMDmitri Moltchanov
Dmitri Moltchanov
SSSarabjot Singh
Sarabjot Singh
YKYevgeni Kourcheyavy
Yevgeni Kourcheyavy
Millimeter-wave (mmWave) systems suffer from a significant loss in performance when the line-of-sight (LoS) path between the transmitter and the receiver is obstructed due to blockage caused by humans or other obstacles. However, due to blocker or user motion, the mmWave receiver may transition between the LoS and non-LoS (nLoS) states. Following the recent 3GPP requirements on spatial consistency for channel modeling, this paper aims to analyze the spatial correlation of blockage statistics and characterize their evolution due to user mobility in a static field of blockers in urban mmWave systems. In particular, we derive conditional probabilities of residing in LoS or nLoS state at a given time $t_1$ , provided that a user was in LoS/nLoS state at a prior time $t_0$ . We demonstrate that for realistic values of user speed, the angle of motion, height of transmitter and receiver, as well as the density of blockers, there always is a significant correlation between user channel states at time $t_0$ and $t_1$ , across the time scales relevant for mmWave resource scheduling. Hence, our model can serve as an important tool for optimizing system performance in the presence of blockage.

Verified
Journal Article
Open access
PDF file
RBRavikumar Balakrishnan
Ravikumar Balakrishnan
MAMustafa Akdeniz
Mustafa Akdeniz
SDSagar Dhakal
Sagar Dhakal
AAArjun Anand
Arjun Anand
AZAriela Zeira
Ariela Zeira
NHNageen Himayat
Nageen Himayat
Client and Internet of Things devices are increasingly equipped with the ability to sense, process, and communicate data with high efficiency. This is resulting in a major shift in machine learning (ML) computation at the network edge. Distributed learning approaches such as federated learning that move ML training to end devices have emerged, promising lower latency and bandwidth costs and enhanced privacy of end users’ data. However, new challenges that arise from the heterogeneous nature of the devices’ communication rates, compute capabilities, and the limited observability of the training data at each device must be addressed. All these factors can significantly affect the training performance in terms of overall accuracy, model fairness, and convergence time. We present compute-communication and data importance-aware resource management schemes optimizing these metrics and evaluate the training performance on benchmark datasets. We also develop a federated meta-learning solution, based on task similarity, that serves as a sample efficient initialization for federated learning, as well as improves model personalization and generalization across non-IID (independent, identically distributed) data. We present experimental results on benchmark federated learning datasets to highlight the performance gains of the proposed methods in comparison to the well-known federated averaging algorithm and its variants.

25 September 2023
Verified
Conference Paper
Closed access
ASAhmet Faruk Saz
Ahmet Faruk Saz
YSYashas Malur Saidutta
Yashas Malur Saidutta
FFFaramarz Fekri
Faramarz Fekri
MAMustafa Akdeniz
Mustafa Akdeniz
BEBrandon Edwards
Brandon Edwards
NHNageen Himayat
Nageen Himayat
We present an attribute-based privacy framework for Federated Learning. Our framework utilizes the Generalized Information Bottleneck (GIB) principle to create functionally compressed representations that obscure designated sensitive attributes from potential inferential adversaries at the server. These functional representations are generated following a minimax adversarial optimization of the privacy and utility bounds on the optimization function. We show that our proposed framework inherently provides attribute-based differential privacy (DP) guarantees, reduces communication overhead and improves utility (e.g., image classification performance). We presented both theoretical and experimental comparisons of our less restrictive attribute-based DP approach with conventional DP.