Artificial Intelligence II

General information

Type:	OPT	Curs:	1	Period:	S semester
ECTS Credits:	3 ECTS

Teaching Staff:

Group	Teacher	Department	Language
Year 1	Jordi Nin Guerrero	Operaciones, Innovación y Data Sciences	ENG

Prerequisites

Cloud Computing
Artificial Intelligence I

Previous Knowledge

- Basic programming skills (Python)
- Basic knowledge on Machine Learning

Workload distribution

- 27 hours of face-to-face lessons (Lessons)
- 15 hours of individual work (Assignments)
- 15 hours of group work (Project)
- 24 hours of independent study (Exam preparation + understanding theory concepts)

COURSE CONTRIBUTION TO PROGRAM

- Deep understanding on Machine Learning methods (supervised learning and unsupervised learning)
- Basic knowledge on Deep Learning as a supervised and unsupervised learning method
- Business evaluation of the different ML techniques

Course Learning Objectives

The main goal of the course is to get a better understanding on how to create Machine Learning models including Deep Learning and apply them in specific business problems. The student will practice how to apply ML with Python creating valid evaluation metrics for a specific business context.

CONTENT

1. Introduction to sklearn This introductory session provides the basic concepts of how to use the Python sklearn library.
2. Advanced sklearn A deep dive into the internals of the sklearn library
3. Understanding decision Trees This session will assume some knowledge about decision trees. Based on this knowledge, this session will introduce concepts such as pruning, balancing and hyperparameters tuning.
4. Gradient boosting Gradient boosting is a machine learning technique for regression, classification and other tasks, which produces a prediction model in the form of an ensemble of weak prediction models, typically decision trees. We will introduce it and explain the main differences between this technique and random forest.
5. Introduction to deep learning Assuming certain knowledge about how the gradient descent works, we will explain the backpropagation method and how it can be use to train a deep neural network.
6. Deep learning for supervised tasks We will introduce the basic concepts to use neural networks for supervised machine learning, such as softmax layers or dropout.
7. Deep learning for unsupervised tasks I This session will introduce autoencoders: a technique used for data compression.
8. Exam

1. Introduction to sklearn

This introductory session provides the basic concepts of how to use the Python sklearn library.

2. Advanced sklearn

A deep dive into the internals of the sklearn library

3. Understanding decision Trees

This session will assume some knowledge about decision trees. Based on this knowledge, this session will introduce concepts such as pruning, balancing and hyperparameters tuning.

4. Gradient boosting

Gradient boosting is a machine learning technique for regression, classification and other tasks, which produces a prediction model in the form of an ensemble of weak prediction models, typically decision trees. We will introduce it and explain the main differences between this technique and random forest.

5. Introduction to deep learning

Assuming certain knowledge about how the gradient descent works, we will explain the backpropagation method and how it can be use to train a deep neural network.

6. Deep learning for supervised tasks

We will introduce the basic concepts to use neural networks for supervised machine learning, such as softmax layers or dropout.

7. Deep learning for unsupervised tasks I

This session will introduce autoencoders: a technique used for data compression.

8. Exam

Methodology

The course combines theoretical lectures with practical sessions. During the sessions, students will solve real problems with their computers using Python and state-of-the-art ML libraries.

ASSESSMENT

ASSESSMENT BREAKDOWN

Description	%
Assignment 1: Sklearn	10
Assignment 2: Decision trees and XGBoost	10
Assignment 3: Deep learning	10
Group project	30
Exam	30
Class attendance and participation	10

Assessment criteria

The evaluation of the course is composed of:

- 10 % Class attendance and participation
- 30% Jupyter notebooks with assignments
- 30 % Group project
- 30 % Final exam

Bibliography

1. Pattern Recognition and Machine Learning. Christopher M. Bishop
2. An Introduction to Statistical Learning: with Applications in R. Gareth James, Daniela Witten, Trevor Hastie, and Robert Tibshirani.
3. Deep Learning. Aaron C. Courville, Ian Goodfellow, and Yoshua Bengio

Timetable and sections

Group	Teacher	Department
Year 1	Jordi Nin Guerrero	Operaciones, Innovación y Data Sciences

Timetable Year 1

From 2023/1/19 to 2023/2/27:
Each Thursday from 10:30 to 12:00.
Each Monday from 14:15 to 15:45. (Except: 2023/1/23, 2023/2/6, 2023/2/13 and 2023/2/20)
Each Thursday from 8:45 to 10:15.

Monday 2023/1/30 from 16:00 to 17:30.

From 2023/2/27 to 2023/3/9:
Each Thursday from 8:45 to 10:00. (Except: 2023/3/2)
Each Monday from 16:00 to 17:30. (Except: 2023/3/6)

Artificial Intelligence II (2225.YR.000565.1)