Understanding what deep learning is and how it works

In the past few years, ML has been the go-to tool for academic research and industries since ML made it possible to learn complicated functions and patterns from highly complex data without human intervention. As early as 1980, theoretical results such as Universal Approximation Theorem seemed to indicate that it may be possible for a neural network to learn any function that existed in a dataset. This is a powerful approach because there are several problems in the real world that traditional methods cannot solve. This led to the birth of DL. Even though DL has been around for about a decade now, it has gotten mainstream attention recently. So, why didn’t DL take off until recently? This can be mainly attributed to the lack of DL frameworks, big data, and efficient hardware to build complex DL models until recently. It’s only been possible to use DL to produce meaningful empirical results due to the introduction of...