Notes From Learning Machine Learning
Machine learning is one of those fields that looks like magic from the outside and like a pile of linear algebra from the inside. As I work toward a master's in ML, here are a few things that have genuinely helped me learn it.
Start with the problem, not the model
It's tempting to reach for the fanciest model first. But the most useful question is almost always: what does the data actually look like, and what am I really trying to predict? A simple model you understand beats a complex one you don't.
Build the intuition by hand
Implementing things like linear regression and a basic classifier from scratch — no libraries — taught me more than any amount of reading. Once you've felt how gradient descent nudges weights step by step, the bigger frameworks stop feeling like black boxes.
Respect the boring parts
Cleaning data, splitting it properly, and choosing the right metric are unglamorous and absolutely critical. Most real-world ML failures aren't exotic — they're a leaky validation set or a mismatched metric.
Keep going
The field moves fast, and that's part of the fun. I'm planning to write up some hands-on projects here as I go.