The teaching is organized as follows:

Learning objectives

Il corso di Advanced Programming for AI fornisce agli studenti una conoscenza approfondita dei linguaggi di programmazione, degli strumenti e delle architetture software nell’ambito dell’intelligenza artificiale, con un’attenzione particolare al deep learning. Il corso è strutturato per potenziare le competenze di programmazione in Python, con un focus su framework fondamentali come TensorFlow, Keras e PyTorch, e su librerie essenziali come Scikit-learn, Pandas, Matplotlib e SciPy, utilizzate per la costruzione di architetture di apprendimento, la visualizzazione dei dati e l’interpretazione dei modelli. Gli studenti acquisiranno un’esperienza pratica sui principali componenti dei sistemi di deep learning, affrontando concetti chiave come funzioni di perdita (loss functions), funzioni di attivazione (activation functions), ottimizzatori (optimizers), tecniche di regolarizzazione (regularization techniques) e normalizzazione del batch (batch normalization). Inoltre, il corso approfondisce metodologie avanzate di deep learning, tra cui meta-learning, transfer learning, domain adaptation, continual learning, active learning, multi-task learning e federated learning, fino alle più recenti innovazioni nel settore, come i Large Language Models e i Visual Language Models.

Prerequisites and basic notions

Students should have a good knowledge of Python, algebra, and machine/deep learning, as this course introduces advanced deep learning topics.

Program

APAI aims to provide information on programming languages, tools, and software architectures emerging in the field of artificial intelligence. Students will improve their Python programming skills, assemble software modules, manage models and patterns, and deploy them on cloud platforms. Special emphasis is placed on key frameworks such as TensorFlow and PyTorch, along with essential libraries such as Scikit-learn, Pandas, Matplotlib, and SciPy, used for building learning architectures, data visualization, and model interpretation. Syllabus: - Introduction to Learning Frameworks: TensorFlow, Keras, PyTorch - Learning Building Blocks: loss functions, activation functions, optimizers, vanishing gradients problems, batch normalization, regularization, dropout - Custom Models: tensors and operations, tensors and NumPy, type conversions, variables, data structures, custom loss functions, custom metrics - Data Loading and Preprocessing: shuffling, parsing, feature preprocessing, transformation chaining - Advanced Methodologies: meta-learning, transfer learning, domain adaptation, continual learning, active learning, knowledge distillation, self-supervised learning, Large Language Models (LLMs), Visual Language Models (VLMs)

Bibliography

Didactic methods

Lectures, laboratory experiments and exercises.

Learning assessment procedures

The assessment is project-based. Students are required to design and implement a project and prepare a technical report describing their work in the format of a scientific paper. At the end of the course, students present and discuss their project during an oral examination. The final grade is based on the quality of the implemented project, the scientific quality and clarity of the written report, and the student's ability to present, justify, and critically discuss the work during the oral examination.

Evaluation criteria

Theoretical and applied knowledge of the techniques taught in the course; critical ability to select techniques based on the problem; ability to use the techniques taught in the course.

Criteria for the composition of the final grade

The final grade will be calculated as the arithmetic mean of the project grade and the oral exam grade.

Exam language

Inglese