I'm new to machine learning and have mostly been learning through YouTube videos. I'm realizing that writing down every detail would create an overwhelming amount of notes. Should I try to memorize common code, such as the syntax for a log transformation, or focus on understanding the underlying logic and look up the exact implementation when I need it? I'd also appreciate advice on how to practice effectively as a beginner.
3 Answers
The best way to learn is to build small projects instead of only watching videos. Practice on datasets through platforms such as Kaggle or similar challenge sites. Try cleaning the data, choosing a model, evaluating it, and explaining your results. When you get stuck, look up the relevant syntax, then close the reference and rewrite the solution yourself so you remember the process.
You don’t need to memorize every line of code. Focus on understanding what each technique does, why you would use it, what assumptions it makes, and how to interpret the result. For common operations, it’s useful to understand the general implementation, but checking documentation for exact syntax is completely normal. Keep notes focused on concepts, workflows, and mistakes to avoid rather than copying entire tutorials.
Books and structured courses can help fill in gaps that short videos often leave behind. Once you understand the fundamentals, read beginner-friendly research papers or practical articles and try reproducing a small part of the method in a project. You don’t need to start with advanced papers; focus on understanding the problem, approach, results, and limitations.
So for something like a log transformation, I should understand why and when it helps, while using documentation to check the exact code when I need it?