Essential for anyone specializing in AI.

Classes

Theory classes are very dense. May be hard to follow if you get lost easily. However, Prof André Martins (only one I attended classes of) is very knowledgeable about the subject and should be able to clarify any questions.

Lab classes are very useful especially if you are a complete noob in the matter. If not, I wouldn't bother as solutions are eventually handed out.

Project

Project can be done safely within deadlines with a consistent group. Grades are seemingly very fair as a correct/working project will get you close to a maximum grade.

Exam

It is very scary....

Theory Classes

Heavy math concepts presented with diferent symbols to represent the same concept in each model. Extremely hard to pay attention to unless you are passionate about deep learning models and calculus.

Practical Classes

Generally good to get you ready for part of the project and the calculations in the mandatory Exam.

Study materials

Enormous amount of slides (plus some repeated ones they use to gradually show stuff during presentations) that aren't even enough to know all the theory questions on the Exam. Previous exams with solutions are provided, which helps.

#...

By far one of the most interesting courses in MEIC. If you plan on doing anything AI-related on your thesis or career this is a no-brainer. Surprisingly modern and up-to-date while giving background and historical context. The exam was extremely difficult but the project, while time consuming, is very doable, and gives you some PyTorch exp.

This course is a must for anyone specializing in Artificial Intelligence or Intelligent Robotics. It covers material ranging from perceptrons all the way to RNNs, CNNs, and transformers.

For the practical evaluation, there are two projects (though they feel more like homework). You complete partially given code, analyze and justify the results, and solve some theoretical exercises on paper. The workload is light, grades are good, but I didn’t find the assignments particularly engaging.

The exam is the challenging part of this course—the syllabus is extensive, and the exercises can vary...

Deep Learning is probably the most important course in the Machine Learning track.

Theoretical lectures

The syllabus is extensive and well-structured, covering a lot of ground without skipping mathematical details (though it's not necessary to understand every single one to follow the course).

Theoretical lectures were taught by André Martins (for MEIC) and Mário Figueiredo (for MEEC). Both are highly knowledgeable and answer student questions clearly and thoroughly. However, neither has a particularly charismatic or engaging teaching style, so the lectures are mostly content-heavy, which...

This feedback is based on the Machine Learning course, since Deep Learning is still new in MEIC. The content appears to follow a similar structure.

The course focuses on neural networks, one of the most widely used machine learning models today. The first part of the course (topics 1 to 3) covers the mathematical foundations of neural networks. Expect a strong emphasis on matrices, derivatives, gradients, and some probability.

The second part of the course dives into different network architectures used to tackle problems in areas like image and text processing. Here, the mathematical depth...

Machine Learning (Aprendizagem) is one of the most important courses for anyone interested in Artificial Intelligence, as it provides a deeper understanding of how ML algorithms actually work.

The course is taught by Andreas Wichert, a well-known and entertaining professor. His lectures often include humor, and he clearly knows the subject deeply.

His research area combines AI with Quantum Computing, and he’s also the author of a concise 200-page book that introduces the course topics well. While it’s expensive (~90€), it's available in the library or online and is worth checking out.

The...