53861 - INTELLIGENZA ARTIFICIALE
Expected Learning Outcomes
According to the Dublin descriptors to the Dublin descriptors, students, at the end of the course, will demonstrate to:
Understand the distinction between paradigms for intelligent systems development based on “classical AI” and on “Machine Learning”
Be able autonomously to design, prototype and evaluate systems that learn from data
Know cognitive services available on Cloud
Be aware of the limitations and of the issues related to the development of intelligent algorithms and systems.
Course Structure
Lectures, practical labs, case study discussion.
Required Prerequisites
Detailed Course Content
Autonomous agents and their environments Cognitive systems architecture (perception, memory, reasoning, action and metacognition) Knowledge representation techniques Reasoning: analogical, case-based, probabilistic. Problem-solving: problem solving “by search” and problem solving “by description”
Classification, regression and predictive models. Supervised learning (decision trees, Support Vector Machines, K-Nearest Neighbors, Neural networks) Unsupervised learning: clustering “Deep learning”: Convolutional neural networks (CNN) and Recurrent Neural networks (RNN). Brief intro to Transformers .“Knowledge discovery from data”: the data mining process, Data visualization and pre-processing, model development and testing, evaluation. Model limitations (explainability, and Bias in dataset) The PyTORCH library for Deep Learning and the Lightning wrapper The RAPID MINER development system
Applications
- Recommender systems
- Sentiment analysis
- Design of conversational agents (chatbot)
Textbook Information
“Grokking Deep Learning”. Andrew W. Trask - Manning Publications (2019), pp. 336.
Selected chapters from:
- “Artificial intelligence: a modern approach”. Stuart Russell, Peter Norvig, 4th edition, 2020. Chap 1 (pp 29-39, 63-93) Chap 2 (pp. 95-140), Chap 3 ( pp. 147-179, 189-199), Cap 24 (pp-856-878).
- “Data science for Business”, O’Reilly media 2013: Chaps 7 & 8 (Decision analytic thinking: what is a good model; and Model Visualization), pp. 187-122.
- “Cognitive computing and big data analytics”. Hurwits et al. Wiley (2013): Cap 9: IBM Watson as a cognitive system, pp. 137-155.
- “Deep Learning With Pytorch.” Eli Stevens, Luca Antiga, - Manning (2019). Chap 2 (pp.15-37) Chap 3 (pp-39-60).
Lecture notes from the instructor (pp.1-60)- Available on STUDIUM
Technical documentation on relevant machine learning libraries and Rapid Miner operators. Available on STUDIUM
Please remember that in compliance with art 171 L22.04.1941, n. 633 and its amendments, it is illegal to copy entire books or journals, only 15% of their content can be copied.
For further information on sanctions and regulations concerning photocopying please refer to the regulations on copyright (Linee Guida sulla Gestione dei Diritti d’Autore) provided by AIDRO - Associazione Italiana per i Diritti di Riproduzione delle opere dell’ingegno (the Italian Association on Copyright).
All the books listed in the programs can be consulted in the Library.
Course Planning
| Subjects | Text References | |
|---|---|---|
| 1 | “Classical” Artificial Intelligence: Reasoning and Problem-solving • Autonomous agents and their environments • Cognitive systems architecture (perception, memory, reasoning, action and metacognition) | R&N - Capitolo 1 (pp 29-39, 63-93) Cap 2 (pp. 95-140), |
| 2 | “Classical” Artificial Intelligence: • Knowledge representation techniques • Reasoning: analogical, case-based, probabilistic. • Problem-solving: problem solving “by search” and problem solving “by description” | R&N - Cap 3 ( pp. 147-179, 189-199) |
| 3 | "Machine learning e data mining": •Classification, regression and predictive models. • Supervised learning (decision trees, Support Vector Machines) • Unsupervised learning: clustering | Dispense del docente (pp.1-60) |
| 4 | ''Machine learning e data mining'': Neural networks | Trask - Capitoli 3,4,5,6 |
| 5 | “Deep learning”: Convolutional neural networks (CNN) and Recurrent Neural networks (RNN). Transformers. | Trask - Capitoli 9,10,11,12 |
| 6 | “Knowledge discovery from data”: the data mining process, model development and testing, evaluation. Data visualization. • Model limitations (explainability, and Bias in dataset) | P&F- Capitoli 7 & 8 (Decision analytic thinking: what is a good model; and Model Visualization), pp. 187-122. Trask - Capitoli 7,8 |
| 7 | The PyTORCH library for Deep Learning. Transformers | S&A - Cap 2 (pp.15-37) Cap 3 (pp-39-60). |
| 8 | Applications and ethical issues: •Recommender systems • Sentiment analysis • Design of conversational agents (chatbot) | B&K&H - Cap 9: IBM Watson as a cognitive system, pp. 137-155. Trask- cap 14 |
Learning Assessment
Examples of frequently asked questions and / or exercises
Versione in italiano