Cours

Nouveau filtre
Nouveau filtre

39032 Cours

Nom complet Nom abrégé Résumé
DATA918 - Apprentissage en ligne et Agrégation TPT_UE_12305

Ce cours constitue une introduction à l’apprentissage en-ligne, c’est-à-dire quand les données sont révélées au fur et à mesure du processus d’apprentissage plutôt que sous la forme d’un échantillon donné une fois pour toutes. Après une rapide introduction aux méthodes incontournables (halving, online gradient), on s’intéressera aux méthodes d’agrégation. L’idée de base est, étant donné plusieurs prédicteurs, de les faire voter en leur attribuant des poids spécifiques plutôt que d’en choisir un seul. Ces méthodes permettront des résultats optimaux dans des conditions extrêmement générales.

Dans un second temps, on reviendra au cadre d’apprentissage “batch” ou “off-line” plus classique: on verra que les méthodes d’agrégation proposées précedemment peuvent également s’utiliser dans ce cas. On discutera également les différents algorithmes possibles pour implémenter ces méthodes: MCMC et méthodes variationnelles.

A la fin du cours, on étudiera des contextes un peu moins standards, comme les problèmes de bandits (où l’on a qu’une information partielle sur l’effet de chaque décision prise) et les problèmes de transfer learning et lifelong learning, qui permettent de réutiliser l’information acquise en résolvant un problème statistique pour en résoudre de nouveaux.

Deux ou trois séances seront dédiées à l’implémentation en R ou Python des algorithmes vus en cours et à leur test sur des jeux de données. Le cours sera évalué par un court projet (résumé d’un article de recherche, implémentation éventuelle des méthodes proposées dans l’article).

Voir le cours
DS-télécom-13 (DATA919) - Computer vision TPT_UE_12306

The ALTEGRAD course aims at providing an overview of state-of-the-art ML and AI methods for text and graph data with a significant focus on applications.

Each session will comprise two hours of lecture followed by two hours of programming sessions.

Grading for the course will be based on a final data challenge.

Voir le cours
DK901 - Social Data Management TPT_UE_2591
 The objective of this class is to present models for the representation of uncertain data, as well as algorithms and tools to process this data, while maintaining information about its uncertainty.
Topics covered include:
● Sources of uncertain data
● Incomplete data models in closed-world assumptions: SQL NULLs and Codd tables, c-tables
● Data model for open-world data: consistent query answering, OBDA
● Possible world semantics
● Querying relational probabilistic databases: operators, lineage, hardness, practical implementations
● Social applications of uncertain data: probabilistic graphs, social influence, crowdsourcing

Labs will feature the MayBMS probabilistic relational database engine.
Voir le cours
DK902 - Natural and Artificial Intelligence TPT_UE_2592
Bringing machines closer to human competence is a fascinating challenge. We can hardly anticipate all the technical consequences that competent machines will have in domains such as human-machine interaction, intelligent search engines, machine translation, robotics, pattern recognition, knowledge mining and learning, adaptive planning or personal assistance.

This course addresses the issue of A.I. as a reverse-engineering problem: try to mimic, not only the performance, but also the processes, of natural intelligence. For example, a text-messaging app reading “The meeting is scheduled for tomorrow.” anticipates future tense: “Will [you be there]?”. It does so through mere statistical association between “tomorrow” and future tense. Could a machine detect that the message is about a future event, and then not only deduce that future tense is appropriate, but also retrieve the reason for attending the meeting?

This course is best adapted to students who want to acquire more than skills in the domain of Artificial Intelligence.
Voir le cours
ECO_5IR01_PD - Economie industrielle TPT_UE_11203 Voir le cours
ECO_5IR03_PD - Économie et management de l’innovation TPT_UE_11204 Voir le cours
ECO_5IR05_PD - Economie des réseaux TPT_UE_11205 Voir le cours
ECO_5IR07_PD - Économie politique des institutions et de la réglementation TPT_UE_11206 Voir le cours
TPT-DATAAI962 - Data Stream Mining TPT_UE_2329
Data streams are everywhere, from F1 racing over electricity networks to social media feeds.
Data stream mining or Real-Time Analytics relies on and develops new incremental algorithms that process streams under strict resource limitations.
This course focuses on, as well as extends the methods implemented in open source tools as MOA and Apache SAMOA.
Students will learn to how select and apply an appropriate method for a given data stream problem; they will learn how to design and implement such algorithms; and they will learn how to evaluate and compare different solutions.
Voir le cours
DK905 - Dynamic Content Management TPT_UE_2594
This module will examine the management of dynamic data, for a variety of distributed Web applications.
The course includes an introduction to standard tools for developing Web applications (REST/SOAP Web Services, XML/JSON, XSLT, BPEL), followed by an exploration of the problems that come from the dynamic nature of the data returned by Web services: wrapper construction, on-the-fly entity resolution, query evaluation using services with limited access patterns, workflow selection, verification/provenance of workflows.
We will also cover the dynamic integration into RDF knowledge bases (Linked Open Data) of the data exported by digital libraries using Web service APIs.

Prerequisites: Basics of the Web (HTTP, HTML, Web forms, XML), Basics of distributed and database systems.
Voir le cours
DK906 - New Data on the Web TPT_UE_2330
This module will teach students the basics of semantic information extraction.
It will cover the concepts, methods, and algorithms to extract factual information from text in order to construct a coherent knowledge base.
This includes some NLP (Part-of-Speech tagging, Dependency Parsing, etc.), and the techniques and concepts of entity disambiguation, instance extraction, the extraction from semi-structured sources (Wrapper Induction, Wikipedia-based approaches), the extraction from unstructured sources (e.g., by Pattern-based approaches), and the extraction by Soft Reasoning (Markov Logic, MAX SAT, etc.).
We will also cover the design of extraction approaches in general (Evaluation, Iteration, etc.), and the alignment of knowledge bases in the Linked Open Data framework.

Propositional & First Order Logic Basics of the Web (HTTP, HTML, (Web forms), XML, ...) Basics of the Semantic Web (knowledge representation, RDF, OWL,...) Graph Theory Java programming
Voir le cours
TPT-DATAAI922 - Big Data Processing TPT_UE_2596
This module will present concepts, architectures and algorithms for IoT big data processing and analytics, at a very large scale, in distributed settings. The following topics will be covered:
    Apache SparkApache FlinkApache Beam/Google Cloud DataFlowApache StormLambda and Kappa Architectures
A strong focus will be given to labs in this class, so that students can gather enough experience with different existing systems, and understand their respective advantages. The architecture of all distributed computing systems will be discussed in detail during lectures.
Voir le cours
ECE_5DA04_TP - Big Graph Databases TPT_UE_2687 Voir le cours
DK909 - New Trends in Data& Knowledge TPT_UE_2597
The Data&Knowledge track acknowledges that new concepts and techniques will be developed over the coming years in the area of knowledge and data mangement.
To ensure the timely coverage of these concepts, and also to welcome potential future lecturers into our track, we allow students to fill the credits of this module completely freely from the courses that are offered at UPSa.
The condition is that the courses be thematically related to knowledge and data management. The organisers of the Data&Knowledge track will examine each proposed course upon request and decide whether to admit it as a possible choice for the students.
Voir le cours
DK910a - Web Data Models TPT_UE_2598
This module will present concepts, architectures and algorithms for data storage, management, and analysis, at a very large scale, especially in distributed settings. The following topics will be covered, each illustrated with a representative system, whose main features will be detailed during lectures:
    Introduction to distributed systems (consistency, availability, and the CAP theorem; ACID vs BASE)Massively distributed (cloud-based) filesystems (e.g., HDFS/GFS)Modern distributed computing: MapReduceDistributed NoSQL databases:
      Dynamic Hash Tables (DHTs)Key-value stores“Big Table” - style systemsGraph databases: Neo4J, PregelDistributed triple storesDocument stores: MongoDB
    Data analysis tools in the Amazon cloud
A strong focus will be given to labs in this class, so that students can gather experience with different existing systems, and understand their respective advantages.
Voir le cours
DK910b - Semantic Web TPT_UE_2709 Voir le cours
DK911a - Data Warehouses TPT_UE_2688
 
Voir le cours
DK911b - Machine Learning TPT_UE_2708 Voir le cours
DK914 - Information Integration TPT_UE_2712 Voir le cours
DK915 - Introduction to Research/Business TPT_UE_2715 Voir le cours
DK916a - Module Liberté - Decision Modeling TPT_UE_2731 Voir le cours
DK916b - Module liberté - Data Camp TPT_UE_2761
 
Voir le cours
DK917 - Factorization-Based Data Analysis TPT_UE_11131 Voir le cours
ECO_5IR09_PD - Économie de la société de l’information TPT_UE_11207 Voir le cours
IREN906 - Economie du Market Design / Economics of market design (delivered in English) TPT_UE_11208 Voir le cours
ECO_5IR11_PD - Management stratégique TPT_UE_11209 Voir le cours
ECO_5IR17_PD - Systèmes d’information et organisation TPT_UE_11211 Voir le cours
ECO_5IR19_PD - Economie de la propriété intellectuelle TPT_UE_11212 Voir le cours
ECO_5IR21_PD - Econométrie avancée TPT_UE_11213 Voir le cours
ECO_5IR26_PD - Transformations numériques - EN2 TPT_UE_12314 Voir le cours