• INFRES+LINCS Seminar:
  • When & Where: Seminars are generally (please, check the timeline and location before coming!) held on:
    • Wednesday afternoon at LINCS (14h-15h, Salle de Conseil) and
    • Thursday afternoon at Barrault (14h-15h, Amphi Saphir).
  • Contact us: if you wish to give a talk on networking, math, software or embedded-system topics, do not hesitate to contact us

[Next talks] [All talks]

11/03/2015Udo Krieger (Otto-Friedrich-University Bamberg)On the Design of Integrated Forwarding andCaching Policies in Named-Data Networking
11/03/2015Patrick Loiseau (EURECOM)Game-theoretic statistics: Learning from data generated by strategic agents
25/03/2015Guillaume Baudart (ENS)A Synchronous View of Loosely Time-Triggered Architectures
15/04/2015Luis Alvarez Corrales (Telecom ParisTech)Analyzing Interference from Static Cellular Cooperation using the Nearest Neighbour Model


Date:11/03/2015, 11h
Room:LINCS, Salle du Conseil
Speaker:Udo Krieger (Otto-Friedrich-University Bamberg)
Talk:On the Design of Integrated Forwarding andCaching Policies in Named-Data Networking
Abstract:Current IP-based networks are unable to fully exploit the capabilities ofthe increasing number of multi-homed mobile terminals. We argue thatNamed-Data Networking (NDN) can fill the gap. First, we elaborate on therequirements for NDN packet forwarding to multi-homed mobile terminalsand categorize forwarding strategies according to their ability to fulfill theserequirements. Further, we describe the initial design of an advanced multi-path forwarding strategy that is derived from performance insights gainedby a CCNx implementation. Secondly, we investigate the performance of aLRU cache replacement policy in the related tree structure. We start withthe performance approximation of Che et al. and assume Zipf-distributedfrequencies of object requests that constitute associated Poisson streams.Then we derive new explicit formulae regarding the hit and miss ratios in thiscascade of LRU caches. Considering adaptive or cooperative cache policies insuch an access network with multi-homed mobile clients, the sketched resultsprovide a basis for an integrated design of efficient forwarding schemes withoptimal cache placement.
Links:back to the list item, list top or summary top

Date:11/03/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Patrick Loiseau (EURECOM)
Talk:Game-theoretic statistics: Learning from data generated by strategic agents
Abstract:Statistical learning methods developed in the last decades have proved very useful for many applications. However, most algorithms were developed under the assumption that the data is independent from the algorithm. This is no longer true in applications where data is generated or provided by (human) strategic agents. As more and more modern applications indeed learn from data provided by external parties, it becomes increasingly crucial to account for the data-provider's incentives in order to design well-performing learning algorithms in practice.In this talk, we use game theory to model explicitly the economic incentives of the learner and of the agents generating data. We show how, by analyzing the game, it is possible to design learning algorithms that work better because they account for the agents' incentives. Specifically, we illustrate this in two examples with applications in security and privacy: (i) Classification of malicious behavior: In this example, a defender runs a classification algorithm to detect malicious behavior while an attacker chooses his attack strategy to balance the utility of the attack and the probability of being detected. We obtain analytic results that show how to optimally perform the classification given the attacker's utility. (ii) Linear regression from agents-provided data: In this example, we study the case of an analyst who tries to infer a model through linear regression from data provided by individuals who add noise to protect their privacy (this technique is called "local privacy"). We again derive results on optimal linear regression and on the price of anarchy of the system that provide methods to better learn from privacy-conscious users by incentivizing them to provide higher-quality data.
Links:back to the list item, list top or summary top

Date:25/03/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Guillaume Baudart (ENS)
Talk:A Synchronous View of Loosely Time-Triggered Architectures
Abstract:Loosely Time-Triggered Architectures (LTTAs) are a proposal forconstructing distributing embedded control systems. They build onquasi-periodic systems, where computing units execute almostperiodically, by adding a thin layer of middleware that facilitatesthe deployment of synchronous applications.In this paper, we show how the deployment of a synchronous applicationon a quasi-periodic architecture can be modeled using a synchronousformalism. Then we detail two LTTA protocols, Back-PressureLTTA reminiscent of elastic circuits, and a new class ofTime-Based LTTA. Finally, we compare the LTTA approach to clocksynchronization protocols which are becoming more an more commonnowadays. Joint work with Albert Benveniste (INRIA)
Links:back to the list item, list top or summary top

Date:15/04/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Luis Alvarez Corrales (Telecom ParisTech)
Talk:Analyzing Interference from Static Cellular Cooperation using the Nearest Neighbour Model
Abstract:The problem of base station cooperation has recently been set within the framework of Stochastic Geometry. Existing works consider that a user dynamically chooses the set of stations that cooperate for his/her service. However, this assumption often does not hold. Cooperation groups could be predefined and static, with nodes connected by fixed infrastructure. To analyse such a potential network, in this work we propose a grouping method based on proximity. It is a variation of the so called Nearest Neighbour Model. We restrict ourselves to the simplest case where only singles and pairs of base stations are allowed to be formed. For this, two new point processes are defined from the dependent thinning of a Poisson Point Process, one for the singles and one for the pairs. Structural characteristics for the two are provided, including their density, Voronoi surface, nearest neighbour, empty space and J-function. We further make use of these results to analyse their interference fields and give explicit formulas to their expected value and their Laplace transform. The results constitute a novel toolbox towards the performance evaluation of networks with static cooperation.
Links:back to the list item, list top or summary top