2 very intense days.
20 hours of building machine learning models. and there is no way on earth I could write a consistent and coherent blog post about that in my current state.
However, after finishing Regression and starting with Classification (Logistic Regression, SVM and Kernel SVM), that involved some heavy math, I came to the conclusion that I need to catch up with some basic math first (probabilities and stats, Linear Algebra, … you name it).
The problem is, I might be a PhD student, but I haven’t read anything math-related since high school. That makes it a bit difficult to catch up in time. I mean, using sklearn, you’d probably get away with no or basic knowledge of math, since you’re using out-of-the-box stuff. So the code is pretty easy, you call a class, adjust a bit, execute and voila! you get the results. You have no idea what happened, but you get results.
So if you know your data and have a good (conceptual) understanding of ML algorithms, it might be okay. for now. even though it’s oddly unsatisfying.
At some point this might not be enough anymore and in order to build more sophisticated and custom tailored models, you just need to know the math. It was quite overwhelming today. As I said, no math since high school. But I have a lot of time. And a lot of ideas. So bring it on.