Source: Deep Learning on Medium
There are other bewildering applications of deep learning in everyday problems, from beating humans at video games, to cloning your voice. It’s good to know that the future of robocalls is bright (if I were you, I’d start investing today). What’s most impressive is that these applications are built on relatively simple mathematical methods.
When I first read through the research papers, I quickly realized that I didn’t fully understand the mathematical foundations that these formulations were built upon. So I decided that when I got back to being a student, I’d take a deep dive into the mathematics of deep learning. And I think I’ve gotten a good start.
I managed to take a student-directed reading course with an amazing applied mathematician here at Dartmouth, Prof. Gelb, along with an applied mathematician, Sriram, and a pure mathematician, Jack. Together, we managed to cover a good chunk of the concepts one would encounter in deep learning: 1st and 2nd order optimization, regression and distribution losses, hyperparameter optimization with a little inference sprinkled in, spectral analysis, and dataset analysis.
Check out the paper! It’s written specifically to cater to a less-mathematical audience, especially because I happen to be a computer scientist, not an applied mathematician.