Machine Learning

Code:

CC4051

Acronym:

CC4051

Level:

400

Keywords
Classification	Keyword
OFICIAL	Computer Science

Instance: 2023/2024 - 2S

Active?	Yes
Responsible unit:	Department of Computer Science
Course/CS Responsible:	Master in Data Science

Cycles of Study/Courses

Acronym	No. of Students	Study Plan	Curricular Years	Credits UCN	Credits ECTS	Contact hours	Total Time
E:BBC	0	PE_Bioinformatics and Computational Biology	1	-	6	42	162
M:A_ASTR	7	Study plan since academic year 2023/2024	1	-	6	42	162
			2
M:CTN	2	Official Study Plan since 2020_M:CTN	1	-	6	42	162
M:DS	22	Official Study Plan since 2018_M:DS	1	-	6	42	162
M:EGEO	3	Official Study Plan.	1	-	6	42	162

Teaching language

English

Objectives

Students should get to know some of the algorithmic, statisticaland computational  foundations of machine learning, as well as concrete methods of machine learning from linear regression to deep and reinforcement learning. They should be able to fundamentedlly select the appropriate algorithms and their hyperparameters for each problem/data set. They should understand and know how to apply methods for inspecting and evaluating approaches and estimating performance.

Learning outcomes and competences

- Understanding the fundamentals of machine learning algorithms and methodologies presented, in particular of deep and reinforement learning.
- Ability to justify the choice of a machine learning solution to a given problem
- Ability to apply the algorithms to new problems
- Ability to evaluate results

Working method

Presencial

Pre-requirements (prior knowledge) and co-requirements (common knowledge)

- Introductory knowledge of data science (for example, obtained in the courses Introduction to Data Science or Data Mining I).
- Programming knowledge preferably in Python. Experienced programmers in other languages should not have any problems.
- Knowledge of data processing with files and in SQL databases
- Knowledge of statistical inference
- Knowledge of basic matrix algebra and calculus in R and R^

Program

In this course we  will (re) visit fundamental concepts and algorithms for model learning and pattern discovery. There will be a focus on their justified application and example-driven experimentation.

Topics:

- Introduction to the area: what is machine learning
- Simple classification and regression models (linear models and nearest neighbor models) and their validation: learning paradigms, loss functions, bias error and variance.
- Foundations and methods for model Evaluation
- Inference methods of models: Search, Expectation-maximization, aggregation.
- Kernel Methods
- Neural networks, deep models and representation learning
- Reinforcement learning
- Unsupervised, semi-supervised and poorly supervised pattern discovery.

Mandatory literature

Hastie Trevor; The elements of statistical learning. ISBN: 0-387-95284-5
Kevin Murphy; Probabilistic Machine Learning: An Introduction, MIT Press, 2022. ISBN: 9780262046824 (Available online: https://probml.github.io/pml-book/book1.html)

Complementary Bibliography

Ian Goodfellow and Yoshua Bengio and Aaron Courville; Deep Learning, MIT Press, 2016

Teaching methods and learning activities

The classes will be partly expositive, with individual and group dynamics involving the students. Practical out-of-class work will be done with classroom support. Students will also be able to do writing and presentation work. There will be a test and a final exam.

Evaluation Type

Distributed evaluation with final exam

Assessment Components

designation	Weight (%)
Exame	30,00
Participação presencial	5,00
Trabalho prático ou de projeto	55,00
Teste	10,00
Total:	100,00

Amount of time allocated to each course unit

designation	Time (hours)
Estudo autónomo	86,00
Frequência das aulas	42,00
Elaboração de projeto	32,00
Total:	160,00

Eligibility for exams

- Active attendance at least 2/3 of classes (unless an exception is granted)
- Grade greater than 10 in the set of practical assignments.
- Score greater than 0 on the test

Calculation formula of final grade

F = min( 0,3*E + 0,1*T +0,55*P  + 0,05*A ; (0,75*E+0,25*T)*1,3 )

F: final grade
E: Exam
T: Test
P: Practical components / assignments
A: Effective Attendance

Exam is mandatory. If the student does not take the exam the final grade is "Fail due to missing exam".

The grade of the test can be overridden by the exam mark. If the test mark T is below the exam mark E, T is replaced by E in the formula.

Special assessment (TE, DA, ...)

Students under special circumstances should describe their situation to the responsible.

Classification improvement

The exam grade can be improved in subsequent seasons to which the student may have access.

The test grade can be improved by the exam grade. 

The assignments grade cannot be improved after submission.

Observations

All the materials of the UC are in moodle.

The materials will be all in English, including the exams. Classes will be taught in English if this is convenient. Students can participate / respond using Portuguese or English.