• 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]

27/05/2015Don TowsleySampling Node Pairs in Large Graphs (Univ. of Massachusetts, Amherst)
27/05/2015Nidhi HegdeLearning WiFi Performance (Alcatel-Lucent Bell Labs)
03/06/2015Eitan Altman (INRIA)Timelines Analysis and competition over popularity, influence and visibility in social networks.
10/06/2015Emilie Kaufmann (INRIA)A novel spectral algorithm for the identification of overlapping communities
17/06/2015Tony Quek (SUTD, Singapore)TBD
01/07/2015Jeremy Leguay (Huawei)Online network optimization with SDN


Date:27/05/2015, 11h
Room:LINCS, Salle du Conseil
Speaker:Don Towsley
Talk:Sampling Node Pairs in Large Graphs (Univ. of Massachusetts, Amherst)
Abstract:Characterizing user pair relationships is important for applications such as friend recommendation and interest targeting in online social networks (OSNs). Due to the large scale nature of such networks, it is infeasible to enumerate all user pairs and so sampling is necessarily used. In this talk I focus on this problem from two perspectives, from an OSN service provider with access to the complete network and from a suspicious BBN summer intern with limited access to the network. Characterizing pair relationships poses a great challenge to an OSN provider even when it possesses the complete topology. The reason is that when sampling techniques, e.g., uniform vertex sampling (UVS) are naively applied, they can introduce large biases, in particular, for estimating similarity distribution statistics of user pairs such as network homophily. Estimating statistics of user pairs is more challenging in the absence of the complete topology information, since an unbiased sampling technique such as UVS is usually not allowed, and exploring the OSN topology is expensive. To address these challenges, we present unbiased sampling methods to characterize user pair properties based on UVS and random walks (RWs) respectively. We evaluate our methods to show their accuracy and efficiency. Finally, we apply our methods to several OSNs and characterize the homophily present in each.
Biography:Don Towsley holds a B.A. in Physics (1971) and a Ph.D. in Computer Science (1975) from University of Texas. He is currently a Distinguished Professor at the University of Massachusetts in the Department of Computer Science. He has held visiting positions at numerous universities and research labs including University of Paris VI, IBM Research, AT&T Research, Microsoft Research, and INRIA. His research interests include networks and performance evaluation.He currently serves as Co-Editor-in-Chief of the new ACM Transactions on Modeling and Performance Evaluation of Computing Systems (ToMPECS). He served as Editor-in-Chief of the IEEE/ACM Transactions on Networking and as an associate editor of numerous journals. He has served as Program Co-chair of INFOCOM 2009, Performance02, and the joint 1992 ACM SIGMETRICS/Performance Conference as well as General Chair of COMSNETS 2012. He is a member of ACM and IEEE.He has received numerous IEEE and ACM awards including the 2007 IEEE Koji Kobayashi Award and several achievement award. He has also received numerous best paper awards including the IEEE Communications Society 1998 William Bennett Paper Award, several test of time awards and conference best paper awards. Last, he has been elected Fellow of both the ACM and IEEE.
Links:back to the list item, list top or summary top

Date:27/05/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Nidhi Hegde
Talk:Learning WiFi Performance (Alcatel-Lucent Bell Labs)
Abstract:Accurate prediction of wireless network performance is important when performing link adaptation or resource allocation. However, the complexity of interference interactions at MAC and PHY layers, as well as the vast variety of possible wireless configurations make it notoriously hard to design explicit performance models.In this work, we advocate an approach of "learning by observation", where we use machine learning techniques to learn implicit performance models, from a limited number of real-world measurements. While our model does not use information on the WiFi mechanism itself, our results show that the accuracy of performance prediction is significantly improved as compared to measurement-seeded models based on SINR. To demonstrate that learned models can be useful in practice, we build a new algorithm that uses such a model as an oracle to jointly allocate spectrum and transmit power. Our algorithm is utility-optimal, distributed, and it produces efficient allocations that significantly improve performance and fairness.Joint work with Julien Herzen (EPFL) and Henrik Lundgren (Technicolor)
Links:back to the list item, list top or summary top

Date:03/06/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Eitan Altman (INRIA)
Talk:Timelines Analysis and competition over popularity, influence and visibility in social networks.
Abstract:The major social networks use timelines to display content that a user receives from those sources that he follows. A content that arrives at a timeline is placed at the top of the list and each other content is pushed down by one step. If the timeline is finite (as in twitter) then this causes the last element in the timeline to be pushed out. Moreover, the lower the content is on the timeline the smaller its influence is. In this talk we shall first present analysis of the timeline process and of the probability to be visible on the timeline. We shall further describe the influence process of content which takes into account not only the location on the timeline but also aging of content. We shall then study the impact of the burstiness of the arrival of contents on the timeline process. We shall finally introduce game theoretical models to describe the competition over popularity and over visibility of content in the timeline. We shall study two specific game theoretic problems. The first is a timing game: when is it best to send a content. The second is a resource allocation game for optimal control of the flow of content.The work describes ongoing collaborative work with many woauthors: Alexandre Reiffers, Nahum Shimkin, Anurag Kumar, Yezekael Hayel and others.
Links:back to the list item, list top or summary top

Date:10/06/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Emilie Kaufmann (INRIA)
Talk:A novel spectral algorithm for the identification of overlapping communities
Abstract:Spectral algorithms are popular methods for finding a partition of a network into groups of nodes that are densely connected, called communities. However, the structure of many real world networks is better explained by a set of overlapping communities than by a partition. In this talk, I will present combinatorial spectral clustering (CSC), a simple spectral algorithm designed to identify overlapping communities. The algorithm is motivated by a random graph model called stochastic blockmodel with overlap (SBMO), under which it is proved to be consistent. Joint work with Thomas Bonald and Marc Lelarge.
Links:back to the list item, list top or summary top

Date:17/06/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Tony Quek (SUTD, Singapore)
Talk:TBD
Abstract:TBD
Links:back to the list item, list top or summary top

Date:01/07/2015, 14h
Room:LINCS, Salle du Conseil
Speaker:Jeremy Leguay (Huawei)
Talk:Online network optimization with SDN
Links:back to the list item, list top or summary top