Graduate School and Research Center in Digital Sciences

Seminar: Workshop on Coordination, Cooperation and Learning in Wireless (and more) Networks

Date: September 4, 2015

Location: Eurecom - Salle des Conseils

2.30pm Welcome 2.45pm A Stochastic Analysis of Network MIMO Systems (Professor Wei Yu, University of Toronto) Abstract: Network MIMO, where base-stations (BSs) cooperatively transmit and receive to/from the users, promises to significantly alleviate the inter-cell interference problem in wireless cellular networks; but its analytical performance characterization is still a difficult open problem. In this talk, we describe a stochastic geometry analysis of a network MIMO system, where the multiple-antenna BSs are distributed according to a Poisson point process and cooperate using zero-forcing beamforming to serve multiple users. We obtain tractable and accurate approximations of the signal power and inter-cluster interference power distributions, and derive a computationally efficient expression for the achievable per-BS ergodic sum rate. The analysis enables us to obtain the optimal number of users to schedule. Further, it allows us to quantify the performance improvement of network MIMO systems as a function of the cooperating cluster size. In particular, due to the zero-forcing penalty across a distributed set of BSs and the inevitable out-of-cluster interference that always exists, the per-BS ergodic sum rate of a network MIMO system does not approach that of an isolated cell even at unrealistically large cluster sizes. Finally, we illustrate the benefit of user-centric clustering for cell-edge users, and remark on a comparison between massive MIMO and network MIMO systems. (Joint work with Raviraj Adve and Kianoush Hosseini) 3.30pm How Well Do We Learn Over Networks? (Professor Ali H. Sayed, UCLA) Abstract: Network science deals with issues related to the aggregation, processing, and diffusion of information over graphs. While interactions among agents can be studied from the perspective of cluster formations, degrees of connectivity, and small-world effects, it is the possibility of having agents interact dynamically with each other, and influence each other's behavior, that opens up a plethora of notable possibilities. For example, examination of how local interactions influence global behavior can lead to a broader understanding of how localized interactions in the social sciences, life sciences, and system sciences influence the evolution of the respective networks. For long, system theory has focused on studying stand-alone dynamic systems with great success. Nowadays, rapid advances in the biological sciences, animal behavior studies, and in the neuroscience of the brain, are revealing the striking power of coordination among networked units. These discoveries are motivating deeper studies of information processing over graphs in various disciplines including signal processing, machine learning, optimization, and control. In this presentation, we examine the learning behavior of adaptive networked agents over both strongly-connected and weakly-connected graphs and describe some interesting patterns of behavior on how information flows over graphs. In the strongly-connected case, all agents are able to learn the desired true state within the same accuracy level even when different agents are subjected to different noise conditions and to different levels of information. In contrast, in the weakly-connected case, a leader-follower relationship develops with some agents dictating the behavior of other agents regardless of the local information clues that are sensed by these other agents. The findings clarify how asymmetries in the exchange of data over graphs can make some agents totally dependent on other agents. This scenario arises, for example, from intruder attacks by malicious agents or from failures by critical links. 4.15pm Privacy in Networks of Interacting Agents (Professor Vincent Poor, Princeton University) Abstract: A number of applications involve networks of distributed agents (e.g., sensors), who must exchange information in order to achieve individual objectives, such as making inferences about their environments. Though such agents can clearly benefit from exchanging information, they may also wish to maintain a degree of privacy in that information exchange. Such situations give rise to a notion of competitive privacy, which can be explored through a combination of information theory and game theory. In particular, information theory can be used to characterize the tradeoff between privacy of data and the usefulness of that data for an individual agent, while game theory can be used to model the interactions between multiple agents each of whom is mindful of that tradeoff. These ideas will be explored in this talk in a general setting, and then particularly in the context of data exchange for distributed state estimation, in which specific solutions can be obtained.

Workshop on Coordination, Cooperation and Learning in Wireless (and more) Networks