PADI was a difficult course consisting of 6 homeworks, 6 labs, a final open-book exam, and an oral discussion of homeworks. The labs were particularly challenging, and we were given no proper material to prepare for them. To make things worse, they were designed assuming we'd use AI tools to complete them, but some professors then didn't like the use of these tools during evaluations, creating an unfair and inconsistent experience. Overall, it was a frustrating course due to this mismatch between preparation and assessment expectations.

Also, in my opinion, homeworks are very useful for the...

Aulas teóricas acessíveis mas não muito interessantes tendo em conta o potencial que esta matéria tinha. A avaliação prática é feita semanalmente em laboratórios de 3h onde temos que resolver uma ficha de exercícios da matéria da semana. Acho essas 3 horas às vezes não suficientes para a ficha, os professores assumem que nós usamos AI caso contrário seria humanamente impossível fazer tanto em tão pouco tempo. O exame é open book portanto aconselho a levar exames anteriores ou exercícios tipo resolvidos para o exame.

Penso que esta cadeira tem um potencial enorme e fica um pouco aquém das expectativas. As aulas de laboratório são demasiado pesadas e estimulam os alunos a usarem ferramentas (como chat e claude) em vez de estimular o raciocínio. Penso que 3 horas para fazer os labs é demasiado pesado. Penso que seria preferível os alunos desenvolverem os projetos ao longo da semana, para perceberem melhor o que estão a realizar. Penso que os slides também poderiam estar um pouco mais explicativos.

I believe this course has room for some improvements when it comes to the support material and resources (they could have provided more exercises and resolutions to help understand the contents, since some previous exams were a bit of a "gamble" in terms of difficulty) and the laboratory classes: the homeworks were accessible and manageable, we were given several days before the deadlines, but the laboratory assessments (which happened in every lab class) were more challenging and, considering the time we had available to complete and submit them, the use of AI was encouraged but criticized at...

This course was a mixed bag. While the workload wasn't heavy and the assignments were easily completed using the available resources, I didn't enjoy the fact that every single lab session served as an assessment. It felt like we were training to work under pressure rather than actually learning how to apply the material. Some professors encouraged the use of AI to solve problems—designing the labs with that reality in mind—while others viewed it as a reason to question your actual understanding of the work; this occasionally hindered the lab sessions themselves, as it made me self-conscious...

I have mixed feelings about this course. On the theorical side, its does a good job on explaining Markov Decision Problems and its variants, and a very bad job on explaining Reenforcement Learning, which is potentialy the most interessant subject and one of the expectations of most who take the class. On the pratical side, the grading is based on the best 3 out of 4 lab+homework. Homework is hand written problem that you have to solve, and lab is essencially completing code on the lab class. Low efford and easy to get good grade but very dissapointing in terms of engagement, I feel likethis...

Overall Impression

I was expecting more from this course. If I had to choose again, I would probably look for a different course. I took it mainly because it's part of the AI specialization.

The course focuses on AI decision-making and how artificial intelligence determines optimal actions, similar to how a smart vacuum cleaner navigates a space.

Labs and Assignments

• Labs are done in groups of 2 and consist of:

  • A homework assignment
  • A Python programming task

• Both components are relatively simple and focus on the same core concepts. -...

The course had a lot of potential but disappointed me.

It talks about a peculiar part of AI (reinforcement learning) but since we have 4 Labs, and the 4th (about reinforcement learning) is optional, it makes so that you do not invest time in the most 'fun' part.

The lab and homework part is really easy to get a good grade.

For the exam, if you invest time in knowing the theoretical part and the practical part of the course, you can do just fine.

Teaching staff

PADI was taught by Professor Francisco Melo, who is hands down one of the best professors I've had at Técnico. His teaching style is super engaging, and he always makes a point of explaining concepts in various ways with easy-to-understand examples, and reviewing material at the beginning of classes.

Practical classes

The practical classes involved implementing code in notebooks in groups of 2, and helped put concepts into practice, although sometimes it felt like throwing mud (or Copilot code) at the wall...

Syllabus

The PADI course is divided into two main topics: Planning and Learning, always having a focus on 'uncertainty' (meaning probabilities).

The main goal of this course is to teach Reinforcement Learning, in which an agent learns through trial and error (e.g. self-driving cars, robots, etc.). All the course content is presented with this goal in mind.

In the first part, we are introduced to Markov chains. They are a way to model probabilistic systems, i.e. systems where there are various states, and the same action in the same state may lead to different outcomes, with a given...