Multimedia Services & Networks

Catalogue des cours de Télécom SudParis

Code

IGFF NET 4252

Niveau

M1

Graduate

Graduate

Semestre

Fall

Domaine

Réseaux

Programme

Programme Ingénieur

Langue

Anglais/English

Crédits ECTS

2,5

Heures programmées

25

Charge de travail

40

Coordonnateur(s)

Département

  • Réseaux et Services Multimédia Mobiles

Organisation

Cours/TD/TP/projet/examen : 6H/6H/6H/6H/1H

Compétences CDIO

  • 1.1 - Connaissance des sciences de base, y compris mathématiques et autres
  • 1.2 - Connaissance des principes fondamentaux d'ingénierie
  • 1.3 - Connaissances avancées en ingénierie : méthodes et outils
  • 2.1 - Raisonnement analytique et résolution de problèmes
  • 2.2 - Méthodes scientifiques : expérimentation, enquête et initiation à la recherche
  • 2.2.1 - Formulation d'hypothèses
  • 2.2.2 - Etat de l'art
  • 2.2.3 - Recherche expérimentale
  • 3.2.3 - Communication écrite
  • 3.2.6 - Présentations orales
  • 4.1.3 - Réglementation de l'ingénierie
  • 4.1.4 - Contexte historique et culturel, culture générale
  • 4.1.5 - Enjeux contemporains

Mots-clés

Radio Mobile Transmisssion & Networks, IoT and AI, AI for Networks, Multimodal AI, Open-Source AI for Networks and Services, Networks for AI, 5G & 6G key technologies, 5G/6G and AI

Contenu

Mobile Radio Transmission and Networks
Applied and pratical AI basic introduction (from the applied and pratical point of view in order to accomplish the two projects of the course)
IoT and IA
IA for Networks & Services
Multimodal IA for Networks & Services
Open-Source IA for Networks & Services
5G/6G key technologies
5G/6G and IA
Networks & Services for IA

Evaluation

grade over 20

Formule de l'évaluation

The evaluation will be based on in-class tests (30%) and a research project (70%).

Approches pédagogiques

This course is divided into two parts: NET 4252 (P1-1) and NET 4254 (P1-2).
This course includes two pratical supervised and guided projects of AI for networks and services.

In the first project, the students will study a use case of AI in network optimization: Q-Learning applied to the control of the transmit power in a LoRa network.
They will simulate an IoT network (LoRa network) in Python and compute its Energy Efficiency (EE) with a communication model.
They will learn how reinforcement learning works (state, action, reward, policy), how Q-Learning improves a table of action values by trial and error, and how exploration and hyperparameters affect whether learning converges.
Finally, they will learn to validate a method by comparing it with a reference and judging the result by its trade-offs.

In the second project, an introduction to Machine Learning and Multimodal AI will be given: How a Machine Learns to See, and How Multimodal AI can connect the gap between images and language. Then, the students will explore the open-source ecosystem, then independently build a video surveillance service powered by a vision-language model.
The students will understand how an intelligent video surveillance service works, and they will learn how to build one using an open-source AI model.
By the end of this second project, the students will be able to:
• Explain what a Machine Learning model is, how it learns, and how to measure its quality.
• Understand how a multimodal model connects images, videos, and text.
• Select, deploy, and utilize an open-source model (Qwen) within a video surveillance service.
• Evaluate an AI service based on concrete criteria: false alarms, latency, cost, and compliance.