Neural networks have been successfully used for computer vision tasks in which machines were previously no match to humans. Surprisingly, some of these networks have a very simple structure. With the help of three concepts that you very likely know or might remember with a little nudge (linear regressions, neurons, and networks) and trusting me on the more complex math that we will skim over, you will learn what makes such neural networks able to handle so much complexity that they can very effectively tell a cat from a dog without ever having pet either. We will also contemplate how much these networks are still able to do even if we start removing their connections, and how that relates to what we understand about their complexity.
Analytics & Operations Management