As a football enthusiast and a beginner in machine learning, the FIFA World Cup was the perfect “first project”. I decided to build a model which would predict the exact number of goals scored by each team in a game.

This is the continuation of my backpropagation post. This time, I will recap how I implemented backpropagation for batch normalization.

Although backpropagation is “just” an application of the (multivariate) chain rule, it took me a while to actually write it down in code. I have seen many sources derive update formulas but what really helped me understand backpropagation was the introduction of computation graphs in Stanford’s CS231n class.

This blog is a collection of my learning notes. Topics include all things related to software engineering, from machine learning to kubernetes to general cloud concepts. The blog is run with Jekyll and Github Pages.