I think many Machine Learning algorithms (and optimization techniques) become clearer when you actually implement them. Writing out the for loops instead of loo

I think many Machine Learning algorithms (and optimization techniques) become clearer when you actually implement them. Writing out the for loops instead of looking at 6 nested summations helped me to understand the runtime complexity of algorithms, and why some things are and aren't feasible to compute. Often it's useful to go start from the "stupidest" implementation and then optimize it. At the end you'll arrive at the formulas that were given to you ;)