When’s the last time you opened up a machine learning paper, saw armies of jargon and mathematics, and decided not to open it? I suspect many people have had this experience, including myself.

Data science is becoming more and more accessible for all. With that, however, comes downsides — many who have taken up machine learning through online resources may have unfamiliarity with reading the technical papers that describe in-depth the very methods they work with every day. The goal of this article is to give those who previously felt like machine learning papers weren’t for them a demonstration of how easy they can be, if armed with the right tools and mindset.

Of course, one article will not be enough to fulfill the hours of university courses and learning that give a solid understanding of the complex mathematics behind machine learning. Hopefully, however, it will be able to give a greater understanding and confidence when reading original machine learning paper breakthroughs. No amount of Googling or online resource can beat the original papers.

We will be walking through the original Batch Normalization paper, considered by many to be one of the greatest breakthroughs in machine learning. This paper, while being full of mathematics, is not overly technical, followable, and intuitive. Along the way, this article will provide breakdowns, tips, notes, and glossaries. Remember to first read the snips from the paper.

#machine-learning #ai #data-science #data-analysis #research