gradient descent and optimization in general I also find extremely cool. I recommend this paper to anyone interested in it