How to Master Gradient Descent Visually

[00:00:00]If there was one thing that every machine learning practitioner had to know, it would be how to minimize a function.

[00:00:08]Whether you're training a neural network, solving logistic regression, or simply trying to teach a computer how to tell a cat from a dog, chances are after a lot of really complex-looking math, your task will boil down to minimizing one gigantic function.

[00:00:25]Geometrically, this means finding the lowest point on the function graph.

[00:00:31]Gradient descent is an extremely popular minimization algorithm.

[00:00:35]It is iterative, which means that instead of finding the solution all at once, it starts at a random location, and then takes repeated steps trying to improve its results.

[00:00:47]Of course, looking at this simple function, this might seem rather unnecessary.