NETWORK INTELLIGENCE

The course aims to provide an in-depth understanding of the principles and methodologies of Network Intelligence (NI), integrating advanced data analytics (Machine Learning, Deep Learning, and Language Models) into wired, wireless, and 5G/6G network architectures.
By leveraging the skills acquired during the course, students will be able to identify the key requirements for a specific application context and to select and design the most appropriate solutions.
By the end of the course, students will be able to identify the best solutions for the design and development of Artificial Intelligence models to be applied to the networking domain.

Probability theory and random variables; deterministic and stochastic signal theory; analog modulation schemes and the main aspects of digital modulation (ASK, FSK, PSK) and channel coding. Fundamentals of statistics and mathematical analysis. Object-Oriented Programming.

Detailed Course Content

1: Introduction: Artificial Intelligence Techniques for Networks (8 hours)

• Basic principles: Machine Learning (ML), Deep Learning (DL) and Modelli Linguistici (LLM)

2: Network Intelligence for signal recognition (10 hours)

• Decoding and synchronization issues in networked systems (sensitivity to fading and multipath); challenges of classical decoding.
• AI-based decoding: the role of Deep Learning (DL) in decoding without synchronization; analysis of the critical trade-off between performance (Bit Error Rate – BER) and model size in resource-constrained systems.
• Chaotic communication systems: Chaos Shift Keying (CSK), chaotic maps (e.g., logistic map and Bernoulli map). Advantages in terms of interference resilience and security; structural pruning techniques based on filter similarity.

3: Network Intelligence for network routing and traffic engineering (10 hours)

• Intelligent routing based on Deep Learning.
• Traffic and congestion forecasting through time-series analysis.
• LLM for the automation and optimization of testing and simulation environments.
• Dynamic adaptation of TCP and QUIC parameters through ML, DL and LLM algorithms.

4: Network Intelligence for protocol design (10 hours)

• Data representation and data harmonization in heterogeneous protocols.
• Use of Modelli Linguistici (LLM) to automate the creation and deployment of southbound interfaces; use of LLM for the automated generation of communication interfaces.
• Virtual Object (VO) for IoT.

5: Network Intelligence for network security (6 hours)

• Social Network Analysis, social structures on online platforms, and analysis of user identities and diffusion patterns.
• Identification and classification of problematic content (e.g., Hate Speech, PCL, Fake News) through AI models (CNN, Transformer).

6: Network Intelligence for Real-time Adaptation in Wireless Networks (10 hours)

• Non-stationarity in 6G systems: dynamic channels, user mobility, and variations in traffic demand; basic concepts of Reinforcement Learning (RL) applied to wireless environments.
• Multi-Armed Bandit models for fast resource allocation under uncertainty: rapid single-step decisions in wireless systems; exploration strategies under variable network conditions; handling non-stationary behavior; deployment of lightweight models in telecommunications testbeds.
• Context-aware bandits: Bandit methods based on simple network indicators.
• Hackathon on Multi-Armed Bandits applied to bandwidth allocation and throughput maximization.

7: Network Intelligence for Network troubleshooting (6 hours)

• Log Analysis and Diagnostic; Knowledge Extraction and diagnosis from log files through LLM.

8: Network Intelligence for Network Control (10 hours)

• Reinforcement Learning frameworks for problems where actions influence the future network states: multi-step effects of decisions in communication networks; state-value and action-value functions; the Bellman Equation; Q-learning to improve long-term performance in wireless systems.
• Training RL algorithms using Gym-based network simulators.
• RL lab for network latency minimization.

9: Deep Reinforcement Learning for Complex 5G Control Problems (5 hours)

• Optimization of wireless control problems via complex agents: from Q-Learning to Deep Q-Learning; handling complex observations in wireless systems; practical stability considerations for Deep Reinforcement Learning under variable wireless conditions.
• Comparison among Multi-Armed Bandit (MAB), Reinforcement Learning (RL), and Deep Reinforcement Learning (DRL) to solve optimization problems in networks.
• Evaluation of DRL algorithms in the ns3 network simulator.
• Deep Reinforcement Learning lab in Open-RAN networks.

10: Coordination and Multi-Agent Decision Making in 6G environments (4 ore)

• The future of AI in 6G networks: multiple network agents simultaneously controlling different parts of the network; coordination problems arising from interference and shared communication resources.
• Multi-agent reinforcement learning for cooperative and competitive scenarios.
• Game theory for stability analysis in multi-agent scenarios.
• Use case: coordination of multiple xApps in O-RAN for the energy saving of Radio Units.

Textbook Information

1. Proakis, J. G., & Salehi, M., Digital Communications, 5th ed., McGraw-Hill, New York, USA.

2.       Luise, M., & Vitetta, G., Teoria dei Segnali, McGraw-Hill Italia, Milano, Italy.

3.       Kurose, J. F., & Ross, K. W., Computer Networking: A Top-Down Approach, 8th ed., Pearson, Boston, USA.

4.       Andreas F. Molich, Wireless Communications: From Fundamentals to Beyond 5G, Wiley, Hoboken, NJ, USA.

5.       Dahlman, E., Parkvall, S., & Sköld, J., 5G NR: The Next Generation Wireless Access Technology, 2nd ed., Academic Press (Elsevier), Cambridge, MA, USA.

6. Lecturer’s notes

Course Planning

	Subject	Text Reference
1	Introduzione: Tecniche di Intelligenza Artificiale per le Reti	1, 3, 6
2	Network Intelligence for signal recognition	2, 6
3	Network Intelligence for network routing and traffic engineering	3, 6
4	Network Intelligence for protocol design	3, 6
5	Network Intelligence for network security	3, 6
6	Network Intelligence for Real-time Adaptation in Wireless Networks	4, 6
7	Network Intelligence for Network troubleshooting	3, 4, 6
8	Network Intelligence for Network Control	3, 6
9	Deep Reinforcement Learning for Complex 5G Control Problems	5, 6
10	Coordination and Multi-Agent Decision Making in 6G environments	6

1.   Proakis, J. G., & Salehi, M., Digital Communications, 5th ed., McGraw-Hill, New York, USA.                                                                                                                                  
2.  Luise, M., & Vitetta, G., Teoria dei Segnali, McGraw-Hill Italia, Milano, Italy.
3.   Kurose, J. F., & Ross, K. W., Computer Networking: A Top-Down Approach, 8th ed., Pearson, Boston, USA.
4.   Andreas F. Molich, Wireless Communications: From Fundamentals to Beyond 5G, Wiley, Hoboken, NJ, USA.
5.   Dahlman, E., Parkvall, S., & Sköld, J., 5G NR: The Next Generation Wireless Access Technology, 2nd ed., Academic Press (Elsevier), Cambridge, MA, USA.

6. Lecturer’s notes

The exam includes two alternative options, to be chosen by the student:

OPTION A — Project + Oral

• Submission of the project by June.

• Oral discussion of the project by the September exam session.

OPTION B — Written + Oral

• Written exam on all the topics covered in the course.

• Oral exam for further discussion following the written test (3 questions).