91250 - Deep Learning

Learning outcomes

At the end of the course, the student understands the foundational ideas, recent advances and application potential of deep neural systems. The student understands supervised and unsupervised techniques, basic neural topologies, methods for visualizing and understanding the behavior on neural nets, adversarial and generative techniques, reinforcement learning, and recurrent networks. The student is able to apply such technologies to solving classification problems in realistic domains.

Course contents

The course begins with an introduction to Neural Networks and Deep Learning, and their typical training mechanism: the backpropagation algorithm.

The main types of neural networks will be discussed: feed forward, convolutional, and recurrent, providing concrete examples and discussing architectures that have proven useful for image processing, localization, segmentation, style transfer, text processing, and many other applications.

Techniques for visualizing the behavior of hidden neural units will be explored (related to deep dreams and inceptionism), as well as techniques for fooling neural networks, and modern generative techniques, including recent diffusion models.

The final part of the course will be dedicated to an introduction to Deep Reinforcement Learning, with particular attention to designing agents for video games, autonomous driving, and other situations that require complex and adaptive intelligent behaviors.

Prerequisites:

The course is based on an integrated course with Machine Learning from the Artificial Intelligence Studies Program. Knowledge of the following is assumed:

• Machine learning
• Analysis
• Algebra
• Python

Readings/Bibliography

Suggested readings:

• Dive into Deep Learning 

• I.Goodfellow, Y.Bengio, A.Courville. Deep Learning. MIT Press.

Specific pointers to on line material will be provided at each lesson, in addition to the slides of the course.

Teaching methods

Frontal lessons based on slides, with discussion of practical examples via pyhton notebooks.

We also foresee laboratories held by tutors, for 12 additional hours.

Assessment methods

Individual project on topics defined by the teacher.

The grade can be optionally integrated by an oral examination.

The assesment method may change in relation with the attendance.

Teaching tools

Lectures will make extensive usage of slides. Working examples will be delivered by means on python notebooks.

Office hours

See the website of Andrea Asperti