University Metropolitan Tirana · Bachelor, Year 3
Machine Learning¶
Building things that work on data nobody cleaned for you.
You already know how to train a model. The hard part is knowing whether to trust it.
Start here Set up your machine
Learn it·Build it·Show it
Fourteen weeks, three team projects, and a portfolio you can send to anyone who asks what you can do.
How a week runs¶
The concept, and the problem it exists to solve.
A short drill together, then straight onto your own project.
Show what you did. Answer questions about it.
The pipeline is the deliverable, not the model.
A score of 95% because you leaked the answer is worth less than 71% you can defend.
Find your way¶
-
Start here
How the course runs, what the 100 points are made of, and how the three projects work.
-
Toolkit
Python, Git, VS Code, and the refreshers. Three hours now saves you the whole semester.
-
Block 1 · Data & Pipeline
Weeks 1 to 4. Framing a problem, data you can trust, features, and honest evaluation.
-
Block 2 · Models & Optimization
Weeks 5 to 9. Linear models, gradient descent, regularization, and why trees keep winning.
-
Block 3 · Representation & Systems
Weeks 10 to 14. Working without labels, PCA, the bridge to neural networks, and shipping.
-
Projects
Three portfolio projects with their specs and rubrics. These are what you show people afterwards.
Where this ends¶
Week 14, in front of the room, defending a system you chose and built.
Start simple, measure honestly, improve in small steps. A clear baseline and good metrics beat a clever algorithm every time.