**Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?**

Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur GraphSearch.

Personne# Soheil Mohajerzefreh

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Unités associées

Chargement

Cours enseignés par cette personne

Chargement

Domaines de recherche associés

Chargement

Publications associées

Chargement

Personnes menant des recherches similaires

Chargement

Domaines de recherche associés

Cours enseignés par cette personne

Aucun résultat

Aucun résultat

Publications associées (5)

Chargement

Chargement

Chargement

Personnes menant des recherches similaires (127)

Unités associées (2)

The main goal in network information theory is to identify fundamental limits of communication over networks, and design solutions which perform close to such limits. After several decades of effort, many important problems still do not have a characterization of achievable performance in terms of a finite dimensional description. Given this discouraging state of affairs, a natural question to ask is whether there are systematic approaches to make progress on these open questions. Recently, there has been significant progress on several open questions by seeking a (provably) approximate characterization for these open questions. The main goal of approximation in network information theory is to obtain a universal approximation gap between the achievable and the optimal performance. This approach consists of four ingredients: simplify the model, obtain optimal solution for the simplified model, translate this optimal scheme and outer bounds back to the original model, and finally bound the gap between what can be achieved using the obtained technique and the outer bound. Using such an approach, recent progress has been made in several problems such as the Gaussian interference channel, Gaussian relay networks, etc. In this thesis, we demonstrate that this approach is not only successful in problems of transmission over noisy networks, but gives the first approximation for a network data compression problem. We use this methodology to (approximately) resolve problems that have been open for several decades. Not only do we give theoretical characterization, but we also develop new coding schemes that are required to satisfy this approximate optimality property. These ideas could give insights into efficient design of future network communication systems. This thesis is split into two main parts. The first part deals with the approximation in lossy network data compression. Here, a lossy data compression problem is approximated by a lossless counterpart problem, where all the bits in the binary expansion of the source above the required distortion have to be losslessly delivered to the destination. In particular, we study the multiple description (MD) problem, based on the multi-level diversity (MLD) coding problem. The symmetric version of the MLD problem is well-studied, and we can directly use it to approximate the symmetric MD problem. We formulate the asymmetric multi-level diversity problem, and solve it for three-description case. The optimal solution for this problem, which will be later used to approximate the asymmetric multiple description problem, is based on jointly compressing of independent sources. In both symmetric and asymmetric cases, we derive inner and outer bounds for the achievable rate region, which together with the gap analysis, provide an approximate solution for the problem. In particular, we resolve the symmetric Gaussian MD problem, which has been open for three decades, to within 1 bit. In the second part, we initiate a study of a Gaussian relay-interference network, in which relay (helper) nodes are to facilitate competing information flows over a wireless network. We focus on a two-stage relay-interference network where there are weak cross-links, causing the networks to behave like a chain of Z Gaussian channels. For these Gaussian ZZ and ZS networks, we establish an approximate characterization of the rate region. The outer bounds to the capacity region are established using genie-aided techniques that yield bounds sharper than the traditional cut-set outer bounds. For the inner bound of the ZZ network, we propose a new interference management scheme, termed interference neutralization, which is implemented using structured lattice codes. This technique allows for over-the-air interference removal, without the transmitters having complete access to the interfering signals. We use insights gained from an exact characterization of the corresponding linear deterministic version of the problem, in order to study the Gaussian network. We resolve the Gaussian relay-interference network to within 2 bits. The new interference management technique (interference neutralization) shows the use of structured lattice codes in the problem. We also consider communication from a source to a destination over a wireless network with the help of a set of authenticated relays, and presence of an adversarial jammer who wishes to disturb communication. We focus on a special diamond network, and show that use of interference suppression (nulling) is crucial to approach the capacity of the network. The exact capacity characterization for the deterministic network, along with an approximate characterization (to within 4 bits) for the Gaussian network is provided. The common theme that binds the diverse network communication problems in this thesis is that of approximate characterization, when exact resolutions are difficult. The approach of focusing on the deterministic/lossless problems underlying the noisy/lossy network communication problems has allowed us to develop new techniques to study these questions. These new techniques might be of independent interest in other network information theory problems.

Suhas Diggavi, Christina Fragouli, Mahdi Jafari Siavoshani, Soheil Mohajerzefreh

We consider the problem of multicasting information from a source to a set of receivers over a network where interme- diate network nodes perform randomized linear network coding operations on the source packets. We propose a channel model for the noncoherent network coding introduced by Koetter and Kschischang in [6], that captures the essence of such a network op- eration, and calculate the capacity as a function of network param- eters. We prove that use of subspace coding is optimal, and show that, in some cases, the capacity-achieving distribution uses sub- spaces of several dimensions, where the employed dimensions de- pend on the packet length. This model and the results also allow us to give guidelines on when subspace coding is beneficial for the pro- posed model and by how much, in comparison to a coding vector approach, from a capacity viewpoint. We extend our results to the case of multiple source multicast that creates a virtual multiple ac- cess channel.

2011Mahdi Cheraghchi Bashi Astaneh, Amin Karbasi, Soheil Mohajerzefreh

Non-adaptive group testing involves grouping arbitrary subsets of $n$ items into different pools. Each pool is then tested and defective items are identified. A fundamental question involves minimizing the number of pools required to identify at most $d$ defective items. Motivated by applications in network tomography, sensor networks and infection propagation, a variation of group testing problems on graphs is formulated. Unlike conventional group testing problems, each group here must conform to the constraints imposed by a graph. For instance, items can be associated with vertices and each pool is any set of nodes that must be path connected. In this paper, a test is associated with a random walk. In this context, conventional group testing corresponds to the special case of a complete graph on $n$ vertices. For interesting classes of graphs a rather surprising result is obtained, namely, that the number of tests required to identify $d$ defective items is substantially similar to what required in conventional group testing problems, where no such constraints on pooling is imposed. Specifically, if $T(n)$ corresponds to the mixing time of the graph $G$, it is shown that with $m=O(d^2T^2(n)\log(n/d))$ non-adaptive tests, one can identify the defective items. Consequently, for the Erd\H{o}s-R'enyi random graph $G(n,p)$, as well as expander graphs with constant spectral gap, it follows that $m=O(d^2\log^3n)$ non-adaptive tests are sufficient to identify $d$ defective items. Next, a specific scenario is considered that arises in network tomography, for which it is shown that $m=O(d^3\log^3n)$ non-adaptive tests are sufficient to identify $d$ defective items. Noisy counterparts of the graph constrained group testing problem are considered, for which parallel results are developed.

2012