• 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, please contact

[Next talks] [All talks]

30/10/2014Jeremy Pitt (Imperial College London)Formal Models of Social Processes: Computational Justice for Fair and SustainableResource Allocation in Socio-Technical Systems.
05/11/2014Richard Combes (Supélec)Unimodal Bandits without Smoothness
12/11/2014Luca De Cicco (Politecnico di Bari)Controlling Adaptive Video Streaming Systems: One Knob is Enough
19/11/2014Bruno Kauffmann (Orange)Catalog Dynamics: Impact of Content Publishing and Perishing on the Performance of a LRU Cache
10/12/2014Marco Mellia (Politecnico di Torino)Is there a case for mobile phone content pre-staging

Date:30/10/2014, 14h
Room:Amphi Saphir, rue Barrault
Speaker:Jeremy Pitt (Imperial College London)
Talk:Formal Models of Social Processes: Computational Justice for Fair and SustainableResource Allocation in Socio-Technical Systems.
Abstract:Many open computing systems, for example grid and cloud computing, and ad hoc networks,such as sensor or vehicular networks, face a similar problem: how to collectivise resources,and distribute them fairly, in the absence of a centralized component. In this talk, we apply themethodology of sociologically-inspired computing, in which the study of human (social) modelsare formalised as the basis of engineering solutions to technical problems. In this case, wepresent formal models of Ostrom's design principles for self-governing institutions andRescher's theory of distributive justice, for defining executable specifications of electronicinstitutions which support fair and sustainable resource allocation in open computing systems.We will also discuss some ramifications of this research: in particular the implications ofunrestricted self-modification of mutable rules for the design of adaptive systems, andthe potential of such systems for reasoning about resource allocation in socio-technicalsystems, where computational intelligence operates on behalf of (or in consort with)human intelligence. This is the basis for a programme of research we callcomputational justice: capturing some notions of `correctness' in the outcomesof algorithmic decision-making as for basis self-governance in socio-technical systems.
Biography:Jeremy Pitt is Reader in Intelligent Systems in the Department of Electrical & Electronic Engineering at Imperial College London, where he is also Deputy Head of the Intelligent Systems & Networks Group. His research interests focus on developing formal models of social processes using computational logic, and their application to multi-agent systems, for example in agent societies, agent communication languages, and self-organising electronic institutions. He also has an strong interest in the social impact of technology, and has edited two recent books, This Pervasive Day (IC Press, 2012) and The Computer After Me (IC Press, 2014). He has been an investigator on more than 30 national and European research projects and has published more than 150 articles in journals and conferences. He is a Senior Member of the ACM, a Fellow of the BCS, and a Fellow of the IET; he is also an Associate Editor of ACM Transactions on Autonomous and Adaptive Systems and an Associate Editor of IEEE Technology and Society Magazine.
Links:back to the list item, list top or summary top

Date:05/11/2014, 14h
Room:LINCS, Salle du Conseil
Speaker:Richard Combes (Supélec)
Talk:Unimodal Bandits without Smoothness
Abstract:In this talk we present new results on the stochastic bandit problems with a continuum set of arms and where the expected reward is a continuous and unimodal function of the arm. Our setting for instance includes the problem considered in (Cope, 2009) and (Yu, 2011). No assumption beyond unimodality is made regarding the smoothness and the structure of the expected reward function. Our first result is an impossiblity result: without knowledge of the smoothness of the reward function, there exists no stochastic equivalent to Kiefer's golden section search (Kiefer, 1953). Further, we propose Stochastic Pentachotomy (SP), an algorithm for which we derive finite-time regret upper bounds. In particular, we show that, for any expected reward function $mu$ that behaves as $mu(x)=mu(x^star)-C|x-x^star|^xi$ locally around its maximizer $x^star$ for some $xi, C>0$, the SP algorithm is order-optimal, i.e., its regret scales as $O(sqrtTlog(T))$ when the time horizon $T$ grows large. This regret scaling is achieved without the knowledge of $xi$ and $C$. Our algorithm is based on asymptotically optimal sequential statistical tests used to successively prune an interval that contains the best arm with high probability. To our knowledge, the SP algorithm constitutes the first sequential arm selection rule that achieves a regret scaling as $O(sqrtT)$ up to a logarithmic factor for non-smooth expected reward functions, as well as for smooth functions with unknown smoothness. This is a joint work with Alexandre Proutière available at : http://arxiv.org/abs/1406.7447
Biography:Bio: Richard Combes received the Engineering degree from Telecom ParisTech (2008), the Master's degree in mathematics from university of Paris VII (2009) and the Ph.D. degree in mathematics from university of Paris VI (2012). He was a visiting scientist in INRIA (2012) and a post-doc in KTH (2013). He is currently an Assistant Professor in Supélec. He received the best paper award at CNSM 2011. His current research interests include communication networks, stochastic systems and their control and machine learning.
Links:back to the list item, list top or summary top

Date:12/11/2014, 14h
Room:LINCS, Salle du Conseil
Speaker:Luca De Cicco (Politecnico di Bari)
Talk:Controlling Adaptive Video Streaming Systems: One Knob is Enough
Abstract:The video part of Internet traffic is booming due to the ever increasingavailability of video content from sitessuch as YouTube (video sharing), Netflix (movie on demand), Livestream(live streaming).Adaptive video streaming systems dynamically change the video contentbitrate and resolution to match thenetwork available bandwidth and user screen resolution.The mainstream approach employed by adaptive streaming systems, employed byNetflix and Youtube, is touse two controllers: one is the stream-switching algorithm that selects thevideo level, the other regulates theplayout buffer length to a set-point.In this talk we show that such an approach is affected by a fundamentaldrawback: the video flows are not able to get thebest possible video quality when competing with TCP flow. We then design anadaptive video streaming system which fixesthis issue by only using one controller. An experimental evaluation showsthe benefits of the proposed control system.
Links:back to the list item, list top or summary top

Date:19/11/2014, 14h - 15h
Room:LINCS, Salle du Conseil
Speaker:Bruno Kauffmann (Orange)
Talk:Catalog Dynamics: Impact of Content Publishing and Perishing on the Performance of a LRU Cache
Abstract:The Internet heavily relies on Content Distribution Networks and transparent caches to cope with the ever-increasing traffic demand of users. Content, however, is essentially versatile: once published at a given time, its popularity vanishes over time. All requests for a given document are then concentrated between the publishing time and an effective perishing time. In this paper, we propose a new model for the arrival of content requests, which takes into account the dynamical nature of the content catalog. Based on two large traffic traces collected on the Orange network, we use the semi-experimental method and determine invariants of the content request process. This allows us to define a simple mathematical model for content requests; by extending the so-called Che approximation, we then compute the performance of a LRU cache fed with such a request process, expressed by its hit ratio. We numerically validate the good accuracy of our model by comparison to trace-based simulation. Joint work with Felipe Olmos, Alain Simonian and Yannick Carlinet
Links:back to the list item, list top or summary top

Date:10/12/2014, 14h
Room:LINCS, Salle du Conseil
Speaker:Marco Mellia (Politecnico di Torino)
Talk:Is there a case for mobile phone content pre-staging
Abstract:Content caching is a fundamental building block of the Internet. Caches are widely deployed at network edges to improve performance for end-users, and to reduce load on web servers and the backbone network. Considering mobile 3G/4G networks, however, the bottleneck is at the access link, where bandwidth is shared among all mobile terminals. Assuch, per-user capacity cannot grow to cope with the traffic demand. Unfortunately, caching policies would not reduce the load on the wireless link which would have to carry multiple copies of the same object that is being downloaded by multiple mobile terminals sharing the same access link.In this paper we investigate if it is worth to push the caching paradigm even farther. We hypothesize a system in which mobile terminals implement a local cache, where popular content can be pushed/pre-staged. This exploits the peculiar broadcast capability of the wireless channels to replicate content "for free" on all terminals, saving the cost of transmitting multiple copies of those popular objects. Relying on a large data set collected from a European mobile carrier, we analyse the content popularity characteristics of mobile traffic, and quantify the benefit that the push-to-mobile system would produce. We found that content pre-staging, by proactively and periodically broadcasting "bundles" of popular objects to devices, allows to both greatly i) improve users' performance and ii) reduce up to 20% (40%) the downloaded volume (number of requests) in optimistic scenarios with a bundle of 100 MB. However, some technical constraints and content characteristics could question the actual gain such system would reach in practice..
Links:back to the list item, list top or summary top