JFLI Seminar December 2014

Date: December 1st, 2014

Time: 14:00-14:45/14:50-15:35

Place: room 007, Faculty of Science Bldg. 7, Hongo Campus, The University of Tokyo


Title: Popularity-based Caching Strategies in Content-Centric Networking

Speaker: Thomas Silverston
Maître de Conférences – Université de Lorraine
JFLI UMI3527 (délégation CNRS)
The University of Tokyo

Information-Centric Networking (ICN) is expected to be at the core of the Future Internet. Indeed, users are only interested by content and there is a need to change the traditional « host-to-host » communication paradigm of the Internet to a « user-to-content » paradigm. Content Centric Networking (CCN) has recently emerged as a promising architecture. It is based on named-data where a packet address names content and not its location. Then, the premise is to cache content on the network nodes and an important feature for CCN is therefore the cache strategy.
In this talk, we will present our caching strategies for CCN. More precisely we will describe MPC, a strategy based on the content popularity and SACS, a strategy based on the popularity of users.
Besides the caching strategy for CCN, we will also emphasize on other issues with CCN such as routing plan or its deployment, and we will advocate the use of Software-Defined Networking (SDN) or the virtualization of network function (NFV) in this context.


Title: Giving sense to point cloud data

Speaker: Joaquim Silvestre
PhD student at Keio SFC university in IkedaLab,
PhD student at ENSAPLV (École nationale supérieur d’architecture Paris la Villette) MAP-MAACC équipe de l’UMR MAP 3495 CNRS/MCC

There is a convergence between the near future advent of Virtual Reality and the progress in large scale 3D scanning methods. While more and more virtual space will be consumed, real space can be more easily captured. For designers it means that new ways of modeling can be set up. In this presentation I discuss about converting and structuring the raw data acquisition in order to use it for algorithmic design process in VR environment.
The data conversion aims to take low-level data types like point clouds or depth maps and convert them in more “conceptual” data types like parametric model or constructed solid geometry. These kind of data can be more easily controlled through code. Before achieving such goal I present here an intermediate step in this paper : understanding the captured space. Computer vision algorithms can easily recognise faces and objects but the structure that hold them together : architecture, is often partially hidden by them. Meanwhile, what is hold together by architectural space give sign on the architecture it self.
By using an initial trial set of captured data from canonical architecture space, I would like to train a machine learning algorithm to fill up missing or obfuscated data by guess.

Laisser un commentaire

Votre adresse de messagerie ne sera pas publiée. Les champs obligatoires sont indiqués avec *