DEEP LEARNINGModulo BASIC
Academic Year 2024/2025 - Docente: SEBASTIANO BATTIATORisultati di apprendimento attesi
The course covers the theory and practice of artificial neural networks, highlighting their relevance both for artificial intelligence applications and for modeling human cognition and brain function. Theoretical discussion of various types of neural networks and learning algorithms is complemented by hands-on practices in the computer lab. Models for classification and regression, as well as neural network architectures (e.g., Deep Learning) will be discussed. The course will present the techniques to design and optimize learning algorithms, and those useful to assess the performance of Machine Learning systems.
Course Structure
The main teaching methods are as follows:
-
Lectures, to provide theoretical and methodological knowledge of the subject;
-
Hands-on exercises, to provide “problem solving” skills and to apply design methodology;
-
Laboratories, to learn and test the usage of related tools.
Required Prerequisites
Basic Calculus anf Math
Algebra and Matrix Notation
Machine learning basic principle
Python programming language
Attendance of Lessons
Strongly reccomended
Detailed Course Content
Linear Models for Regression: Linear Models for Classification: Gradient Descent, Multi-Class Classification, Classifiers Evaluation
Neural models and Network Architectures
Basic neural network models: multilayer perceptron, distance or similarity based neural networks, associative memory and self-organizing feature map, radial basis function based multilayer perceptron, neural network decision trees, etc.
Basic learning algorithms: the delta learning rule, the back propagation algorithm, self-organization learning, etc.
Supervised Learning with Neural Networks
Deep Learning: Convolutional Neural Network
Textbook Information
DEEP LEARNING FROM BASICS TO PRACTICE (2020)
https://www.glassner.com/portfolio/deep-learning-from-basics-to-practice/
Dive into Deep Learning (2020)
Understanding Deep Learning by Simon J. D. Prince Hardcover ISBN: 9780262048644 Pub date: December 5, 2023 Publisher: The MIT Press 544 pp., 8 x 9 in, 268 color illus., 15 b&w illus.
OTHER
E. Alpaydin, “Introduction to Machine Learning”, MIT Press, 2014
I. Goodfellow, Y. Bengio and A. Courville, "Deep Learning", MIT Press, 2016
M. P. Deisenroth, A A. Faisal, and C. Soon On, Mathematics for Machine Learning, MIT Press, 2019
Course Planning
| Subjects | Text References | |
|---|---|---|
| 1 | Logistic Regression | Glassner (vol .1, vol 2) |
| 2 | BacKpropagation | Glassner (vol .1, vol 2) |
| 3 | Supervised vs Unsupervised Learning | Glassner (vol .1, vol 2) |
| 4 | Neural Network principles | Bishop |
| 5 | Convolutional Neural Networks | Dive into Deep Learning |
Learning Assessment
Learning Assessment Procedures
Writtten and Oral Examination
The test is structured so that each student is given a grade according to the following scheme:
- Not approved: the student has not acquired the basic concepts and is not able to answer at least 60% of the questions or carry out the exercises.
- 18-23: the student demonstrates minimal mastery of the basic concepts, his content connection skills are modest, he is able to solve simple exercises.
- 24-27: the student demonstrates good mastery of the course contents, his skills in connecting the contents are good, he solves the exercises with few errors.
- 28-30 cum laude (distinction): the student has acquired all the contents of the course and is able to master them completely and connect them with a critical spirit; solves the exercises completely and without errors.
Students with disabilities and/or DSA must contact the teacher and the DMI CInAP contact person sufficiently in advance of the exam date
to communicate that they intend to take the exam taking advantage of the appropriate compensatory measures.Examples of frequently asked questions and / or exercises
Example of Algorithms based on training data
Cross Validation
NN Architecutre