The course aims to introduce advanced technologies of artificial intelligence, such as agent and multi-agent systems, probability-based reasoning, and the fundamentals of logic-based reasoning and decision support systems.
Developments in recent artificial intelligence models and technologies and their applications in the domains of major interest, such as robotics, AI-powered assistants, AI in education, finance, health-care and gaming, will be discussed.
The main and recent ethical, social and epistemological issues associated with the use of large-scale artificial intelligence tools and generative Artificial general intelligence will be discussed.
Developments in recent artificial intelligence models and technologies and their applications in the domains of major interest, such as robotics, AI-powered assistants, AI in education, finance, health-care and gaming, will be discussed.
The main and recent ethical, social and epistemological issues associated with the use of large-scale artificial intelligence tools and generative Artificial general intelligence will be discussed.
Curriculum
teacher profile teaching materials
Core Documentation
Lecture notes by the professor.Reference Bibliography
S.J.Russel, P.Norvig "Artificial Intelligence: A Modern Approach", 4/Ed (2020). Pearson Education.Type of delivery of the course
Lessons and exercisesAttendance
Attendance is not compulsory, but it is strongly recommended.Type of evaluation
Written test. teacher profile teaching materials
Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Lecture notes provided by the teacher.
Programme
Foundations of symbolic Artificial Intelligence and problem solving through search in state spaces: problem formulation, uninformed and informed search algorithms, heuristic design, and metaheuristics (hill climbing, simulated annealing, genetic algorithms). Adversarial games: minimax and alpha-beta pruning. Brief overview of rule-based systems and knowledge representation.Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Core Documentation
Stuart Russell, Peter Norvig: Artificial Intelligence: A Modern Approach – Pearson Education.Lecture notes provided by the teacher.
Attendance
Not mandatory, but strongly recommended.Type of evaluation
The assessment of learning consists of a written exam and a practical exam. The written exam evaluates the understanding of the theoretical foundations of Artificial Intelligence, search and optimization algorithms, and generative models. The practical exam assesses the ability to apply knowledge to real-world problems through the design and/or implementation of AI algorithms or models. The final grade is determined based on the overall evaluation of both exams. teacher profile teaching materials
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Core Documentation
Lecture notes by the professor.Type of delivery of the course
Lessons and exercisesAttendance
Attendance is not compulsory, but it is strongly recommended.Type of evaluation
Written test. teacher profile teaching materials
Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Lecture notes provided by the teacher.
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Programme
Foundations of symbolic Artificial Intelligence and problem solving through search in state spaces: problem formulation, uninformed and informed search algorithms, heuristic design, and metaheuristics (hill climbing, simulated annealing, genetic algorithms). Adversarial games: minimax and alpha-beta pruning. Brief overview of rule-based systems and knowledge representation.Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Core Documentation
Stuart Russell, Peter Norvig: Artificial Intelligence: A Modern Approach – Pearson Education.Lecture notes provided by the teacher.
Attendance
Not mandatory, but strongly recommended.Type of evaluation
The assessment of learning consists of a written exam and a practical exam. The written exam evaluates the understanding of the theoretical foundations of Artificial Intelligence, search and optimization algorithms, and generative models. The practical exam assesses the ability to apply knowledge to real-world problems through the design and/or implementation of AI algorithms or models. The final grade is determined based on the overall evaluation of both exams.Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
teacher profile teaching materials
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Core Documentation
Lecture notes by the professor.Type of delivery of the course
Lessons and exercisesAttendance
Attendance is not compulsory, but it is strongly recommended.Type of evaluation
Written test. teacher profile teaching materials
Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Lecture notes provided by the teacher.
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Programme
Foundations of symbolic Artificial Intelligence and problem solving through search in state spaces: problem formulation, uninformed and informed search algorithms, heuristic design, and metaheuristics (hill climbing, simulated annealing, genetic algorithms). Adversarial games: minimax and alpha-beta pruning. Brief overview of rule-based systems and knowledge representation.Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Core Documentation
Stuart Russell, Peter Norvig: Artificial Intelligence: A Modern Approach – Pearson Education.Lecture notes provided by the teacher.
Attendance
Not mandatory, but strongly recommended.Type of evaluation
The assessment of learning consists of a written exam and a practical exam. The written exam evaluates the understanding of the theoretical foundations of Artificial Intelligence, search and optimization algorithms, and generative models. The practical exam assesses the ability to apply knowledge to real-world problems through the design and/or implementation of AI algorithms or models. The final grade is determined based on the overall evaluation of both exams.Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
teacher profile teaching materials
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Core Documentation
Lecture notes by the professor.Type of delivery of the course
Lessons and exercisesAttendance
Attendance is not compulsory, but it is strongly recommended.Type of evaluation
Written test. teacher profile teaching materials
Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Lecture notes provided by the teacher.
Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
Programme
Foundations of symbolic Artificial Intelligence and problem solving through search in state spaces: problem formulation, uninformed and informed search algorithms, heuristic design, and metaheuristics (hill climbing, simulated annealing, genetic algorithms). Adversarial games: minimax and alpha-beta pruning. Brief overview of rule-based systems and knowledge representation.Foundations of NLP and neural networks for sequence modeling (embeddings, RNNs, LSTMs) with an introduction to the attention mechanism. Transformer architecture: self-attention, multi-head attention, positional encoding (including RoPE), encoder-decoder and decoder-only architectures.
Large Language Models (LLMs): architecture, prompting, in-context learning, and sampling strategies. Training and adaptation (pretraining, fine-tuning, LoRA, quantization). Alignment methods (RLHF, DPO) and brief introduction to reasoning.
Agentic systems and RAG: retrieval-augmented generation, function calling, and agent frameworks (ReAct). Evaluation of models (LLM-as-a-judge), limitations, biases, and main research directions.
Core Documentation
Stuart Russell, Peter Norvig: Artificial Intelligence: A Modern Approach – Pearson Education.Lecture notes provided by the teacher.
Attendance
Not mandatory, but strongly recommended.Type of evaluation
The assessment of learning consists of a written exam and a practical exam. The written exam evaluates the understanding of the theoretical foundations of Artificial Intelligence, search and optimization algorithms, and generative models. The practical exam assesses the ability to apply knowledge to real-world problems through the design and/or implementation of AI algorithms or models. The final grade is determined based on the overall evaluation of both exams.Mutuazione: 20801730 INTELLIGENZA ARTIFICIALE in Ingegneria informatica e dell'intelligenza artificiale LM-32 RUSSO PAOLO, LIMONGELLI CARLA,
