Le programme détaillé : Programme-IA2-2019
Cours Fondamentaux (3h)
- Time Series Analytics for IoT and Smart Cities
Themis Palpanas (Université Paris Descartes, LIPADE)
Abstract:
There is an increasingly pressing need, by several applications in diverse domains, for developing techniques able to manage and analyze very large collections of sequences, or data series. Several of the examples of such applications come from IoT and smart cities applications, as well as from a multitude of scientific domains that need to apply machine learning techniques for knowledge extraction. It is not unusual for these applications to involve numbers of data series in the order of hundreds of millions to billions, which are often times not analyzed in their full detail due to their sheer size. However, no existing data management solution (such as relational databases, column stores, array databases, and time series management systems) can offer native support for sequences and the corresponding operators necessary for complex analytics.
In this seminar, we discuss the basics of data series management (distance measures, summarized representations) and analysis (clustering, classification, motifs, discords), and we describe recent efforts in designing techniques for indexing and analyzing data series collections that enable scientists to run complex analytics on their data. Finally, we present open research problems in the area of big sequence management, including promising directions in terms of storage, distributed processing, and query benchmarks.
Short bio:
Themis Palpanas is Senior Member of the French University Institute (IUF), a distinction that recognizes excellence across all academic disciplines, and professor of computer science at the Paris Descartes University (France), where he is director of diNo, the data management group. He received the BS degree from the National Technical University of Athens, Greece, and the MSc and PhD degrees from the University of Toronto, Canada. He has previously held positions at the University of California at Riverside, University of Trento, and at IBM T.J. Watson Research Center, and visited Microsoft Research, and the IBM Almaden Research Center.
His interests include problems related to data science (big data analytics and machine learning applications). He is the author of nine US patents, three of which have been implemented in world-leading commercial data management products. He is the recipient of three Best Paper awards, and the IBM Shared University Research (SUR) Award.
He is currently serving on the VLDB Endowment Board of Trustees, as an Editor in Chief for the BDR Journal, Associate Editor in the TKDE, and IDA journals, as well as on the Editorial Advisory Board of the IS journal, and the Editorial Board of the TLDKS Journal. He has served as General Chair for VLDB 2013, Associate Editor for VLDB 2019 and 2017, Research PC Vice Chair for ICDE 2020, and Workshop Chair for EDBT 2016, ADBIS 2013, and ADBIS 2014, General Chair for the PDA@IOT International Workshop (in conjunction with VLDB 2014), and General Chair for the Event Processing Symposium 2009.
- Apprentissage par renforcement
Philippe Preux (Université de Lille, CRIStAL)
Abstract:
Reinforcement learning is a domain of machine learning. RL belongs to the field of sequential decision making under uncertainty. The RL problem is for an agent to learn to behave in order to solve a certain task (learning to play a game, learning to speak in natural language, learning to manage resources, learning to control a robot, …). During this tutorial, I will present the reinforcement learning problem, formalize it, present the main algorithms to solve it, illustrate it on use cases, explain some best practices to use RL to solve your problem.
Short bio:
Philippe Preux is a professor in Computer Science at the Université de Lille since 2003. He is a member of UMR CRIStAL and Inria-Lille. He has been doing research in AI since around 1992, first about genetic algorithms, then slowly drifting towards machine learning and reinforcement learning. He has created the research group SequeL at Inria in 2006, focusing on sequential decision making under uncertainty. His research is mostly about the design of learning algorithms and their applications.
- Raisonnement spatio-temporel pour l’environnement
Florence Le Ber (Université de. Strasbourg/ENGEES, ICube)
Le cours présentera différents formalismes, les modèles (algèbres et treillis) sous-jacents, les problèmes posés et quelques éléments sur leur résolution. Des extensions développées pour la géomatique et l’analyse d’images seront présentées. Des exemples d’application à l’agriculture et l’environnement seront utilisés comme illustration.
- Data Stream Mining
Joao Gama (University of Porto)
Abstract:
Data Mining is faced with new challenges. Large volumes of data are being generated from real time application domains such as mobility networks, social media, TCP/IP traffic, sensor networks, telecommunication networks etc. The challenge of deriving insights from the Internet of Things (IoT) has been recognized as one of the most exciting and key opportunities for both academia and industry. Advanced analysis of big data streams from sensors and devices is bound to become a key area of data mining research as the number of applications requiring such processing increases. Dealing with the evolution over time of such data streams, i.e., with concepts that drift or change completely, is one of the core issues in IoT stream mining. This tutorial is a gentle introduction to mining IoT big data streams. We introduce data stream learners for classification, regression, clustering, and frequent pattern mining.
Short bio:
João Gama is an Associate Professor at the University of Porto, Portugal. He is a senior researcher and member of the board of directors of the LIAAD, a group belonging to INESC Porto. He is Director of the master Data Analytics. He serves as member of the Editorial Board of MLJ, DAMI, TKDE, NGC, KAIS, and IDA. He served as Chair of ECMLPKDD 2005 and 2015, DS09, ADMA09 and a series of Workshops on KDDS and Knowledge Discovery from Sensor Data with ACM SIGKDD. His main research interest is in knowledge discovery from data streams and evolving data. He is the author of a monograph on Knowledge Discovery from Data Streams. He has extensive publications in the area of data stream learning.
- Optimisation sous contraintes appliquée aux villes intelligentes
Christine Solnon (INSA Lyon, LIRIS) et Hervé Rivano (INSA Lyon, CITI Lab)
– Nous décrirons tout d’abord les principales approches utilisées pour prévoir les conditions de circulation à partir de données issues de capteurs, et nous comparerons ces approches sur deux jeux de données réels (provenant des villes de Lyon et de Marseille).
– Nous introduirons ensuite différentes approches pour optimiser les tournées de livraison en ville dans le cas où nous disposons de prévisions sur les conditions de trafic, et nous évaluerons l’intérêt de ces approches par rapport à des approches considérant que les conditions de trafic sont toujours les mêmes quelle que soit l’heure de départ.
– Enfin, nous introduirons des approches permettant d’optimiser des tournées dans un contexte où les données sont incertaines (du fait qu’elles proviennent de modèles prédictifs), et nous illustrerons l’intérêt de ces approches sur deux applications réelles.
Cours Avancés (1h30)
- Analyse des réseaux – Application à l’analyse de la structure et de dynamique des villes
Marc Barthelemy (CEA, IPhT)
His research interests moved towards applications of statistical physics to complex systems, complex networks, theoretical epidemiology, and more recently on spatial networks. Focusing on both data analysis and modeling with the tools of statistical physics, he is currently also working on various aspects of the emerging science of cities. Marc Barthelemy authored more than 120 papers in international journals, is a co-author of « Dynamical processes in complex networks » (Cambridge Univ. Press, 2008), and authored the two books « Structure and dynamics of cities » (Cambridge Univ. Press, 2016) and « Morphogenesis of spatial networks » (Springer 2018).
- Multiple Aspect Trajectory Mining
Karine Zeitouni (Université de Versailles St Quentin – DAVID)
Abstract:
The pervasiveness of smartphones and various connected sensors and wearables is leading to a wide collection of Multiple Aspect Trajectories, fueling various applications and services. Trajectories capture the movement of humans, vehicles, robots, and phenomena. Nowadays, pure movement data are enriched with multiple heterogeneous contextual aspects either derived from external sources (e.g.., POIs, localized events), or from connected devices and lifelogging. Trajectories can also be transformed to semantic episodes in different ways. As such, mining trajectories can take various forms depending on the considered aspect, the application goal and the environment (indoor, outdoor, surrounding constraints, etc.). In this talk, I will introduce the topic of trajectory mining, and give an overview of the state of the art. In the second part of this talk, I will focus on the methods making use of contextual dimension together with the spatio-temporal aspect. We will conclude with the remaining challenges in trajectory data mining in the context of the emerging Internet of Mobile Things (IoMT).
Short bio:
Karine Zeitouni is a Professor in Computer Science at the University of Versailles Saint-Quentin. She is heading the Ambient Data Access and Mining (ADAM) group at DAVID laboratory. Her main research interest lies in spatial and/or temporal databases and data mining, with a focus on applications in the fields of transportation, environment, and universe science. She is involved in several projects funded by national grants, bilateral research collaborations, and by the European H2020 research and innovation programme. She is a general co-chair of the IEEE Mobile Data Management Conference (MDM 2020), TPC chair of the International Conference on Big Data Research (ICBDR 2019) and TPC Chair of the French conference on data management (BDA 2019).
- Partage de ressources
Sylvain Bouveret (Université de Grenoble, LIG)
Abstract:
La situation dans laquelle un certain nombre d’individus
doivent partager une ressource commune se retrouve dans d’innombrables
exemples de la vie quotidienne, et est également omniprésente dans le
contexte de la ville intelligente. Cette ressource peut être constituée
d’un ensemble d’objets, ou encore d’une ressource infiniment divisible
telle qu’une réserve d’eau, un réseau d’électricité, ou encore du temps
d’utilisation d’un équipement commun. Dans cet exposé, nous aborderons
la problématique du partage équitable de ressources à partir d’exemples
concrets. Puis, nous présenterons les différentes manières de modéliser
et de résoudre ce problème, à travers le prisme de l’informatique et de
l’intelligence artificielle. Cela nous amènera à présenter différents
modèles d’équité, à aborder le sujet de la représentation des
préférences dans ce contexte, et à évoquer les aspects liés à
l’algorithmique et à la complexité du problème.
Short bio:
Diplômé de l’École Nationale Supérieure de l’Aéronautique et de
l’Espace et de l’Université Paul Sabatier de Toulouse en 2004, Sylvain
Bouveret a ensuite effectué une thèse en Informatique à l’Université de
Toulouse, portant sur les aspects algorithmiques et la complexité des
problèmes de partage équitable. Après avoir soutenu cette thèse en
2007, il a travaillé pendant quatre ans en tant qu’ingénieur de
recherche à l’Onera Toulouse. Il est depuis 2011 Maître de Conférences
en Informatique au Laboratoire d’Informatique de Grenoble, Ensimag,
Université Grenoble-Alpes. Sylvain Bouveret effectue ses recherches principalement dans le domaine
de l’intelligence artificielle et du choix social computationnel, avec
une spécialisation sur les problèmes de partage équitable, le vote, et
la représentation de préférences. Bien que ses domaines techniques de
prédilection soient l’algorithmique et la complexité, il a récemment
étendu ses activités de recherche à la visualisation de données et aux
expérimentations dans le domaine du vote.
- Approches Multi-agents pour le transport intelligent
Olivier Simonin (INSA Lyon, CITI Lab)
- A short introduction into deep learning … and when to learn (and when not)
Christian Wolf (INSA Lyon, LIRIS)
Abstract:
In this talk we will give a (necessarily very short) introduction into deep learning, i.e. learning hierarchical high-capacity models from large amounts of data. After a short explanation of deep networks, gradient backpropagation and its implementation through auto-grad in a standard deep learning framework, the talk will focus on two points: (i) the visualization and transfert from learned knowledge from source data to a target application, including efforts to model the shift in distribution, and (ii) the combination of deep learning with more traditional models in the context of robot and vehicle navigation.
Short bio:
Christian WOLF is associate professor (Maître de Conférences, HDR) at INSA de Lyon and LIRIS, a CNRS laboratory, since sept. 2005. He is interested in machine learning and computer vision, especially the visual analysis of complex scenes in motion. His work puts an emphasis on modelling complex interactions of a large amount of variables: deep learning, structured models, and graphical models, and more recently the connections between machine learning and control.
He received his MSc in computer science from TU Vienna, Austria, in 2000, and a PhD in computer science from INSA de Lyon, France, in 2003. In 2012 he obtained the habilitation diploma, also from INSA de Lyon. Between September 2017 and August 2019 he was on leave at INRIA, at the chroma work group at the CITI laboratory. He is currently the national project coordinator of ANR/NSERC project « Deepvision » “Learning highly complex visual problems with deep structured models” (9/2016-3/2020) and the national coordinator of project ANR Delicio « Data and Prior, Machine Learning and Control » (2019-2023); he has coordinated the INSA-Lyon partner of ANR Canada (2007-2010) and PIA Interabot (2012-2016). He is member of the scientific committee of GDR IA and member of the directing committee of GDR ISIS and co-leader of the topic « Machine Learning ». He has supervised 10 defended PhD theses.