The course covers both machine learning and classical AI paradigms in a network context. Here, the concept of network is understood in a broad scientific sense (e.g. networks of agents, sensors, computing nodes, social networks or communication networks).
The course covers theories, methods and practical applications of artificial intelligence in network environments. The focus is on both machine learning and other AI paradigms such as search, planning, reasoning and multi-agent systems. The course is divided into three parts:
- Part I: Basic concepts and tools
Introduction to networked AI, intelligent agents and multi-agent systems. Statistical foundations and optimization methods for distributed systems. Classical AI methods for search, problem solving, knowledge representation and logical reasoning in networked environments. - Part II: Distributed learning and decision making
Optimization over networks and consensus algorithms. Multi-agent systems with communication, coordination and negotiation. Game theory and strategic decision-making in networks. Planning under uncertainty, including Markov decision processes (MDP), distributed MDPs, and partially observable MDPs (POMDPs). - Part III: Advanced topics and applications
Federated and distributed learning, including communication and security aspects. Neurosymbolic AI and the combination of logic and machine learning. Practical applications in areas such as smart grids, the Internet of Things, autonomous vehicles, and wireless networks.
The course includes both theoretical lectures and exercises. The exercises consist of mathematical problems, analysis of algorithm properties, and practical implementations in Python or a similar programming environment.