On a more personal note, I must say that a relatively newly found affinity and aptitude for math was one of the main reasons I decided to take the plunge and make the shift towards a computer science post baccalaureate degree in the first place. I knew in my heart of hearts that a field involving some more axiomatic, rigorous and propositional logic, a field that draws from linear algebra and calculus would be easier for me to invest time and energy in than the health career track I had set myself up for since high school.
I know that whatever entry level job I find myself starting in or even more of the jobs that populate the promotion latter will likely not draw too much on rigorous, theoretical calculation, but it always makes me excited to see the graphs, the calculus behind the time complexity calculations, and now, the calculus behind neural nets. This week, I spent a lot of time exploring the more fundamental question, ‘What really is a neural net?’ Thankfully, Grant Sanderson from 3blue1brown posted a spectacularly thought-provoking and spiffily animated four part video series on the topic.
Fortunately, I find myself also taking multivariable calculus at a community college after what feels like a long hiatus from taking in-person classes. It feels so charming to be seated in a plastic chair again notebook in hand. But I digress. The point is that I will be more primed to revisit and mull over the equations behind neural nets, particularly the involvement of the generalized chain rule and the calculation of gradients.
These are nitty gritty details of math that something like Tensorflow library and the Keras API abstract away quite a bit and, though it’s not strictly necessary for me to understand the math down to the fundamentals to build a model, it is always nice to be able to peak under the hood. Maybe more importantly, this video series and these questions have given me my best exposure yet to what may lie ahead if I choose to pursue a graduate degree in machine learning.
These are the sort of videos whose conceptual density warrant a rewatch or two, but if there’s something worth getting my brain scrambled over, it is most certainly this. After all, 3blue1brown’s most common tagline is ‘pause and ponder’. I will most certainly be hitting pause frequently to fill in some time to ponder. Here’s a link to the video playlist:
