info This post is incomplete. I doubt I’ll ever get around to finishing it :/ Oh well!
For a long time, neural networks seemed too complicated, too fantastical, to ever understand. After I sat down and read through everything, it dawned on me that the whole arrangement, even the particularly wild networks, were just math. Duh.
Here, I’d like to walk through two of the most basic neural networks, one with only dense layers and the other with some convolutional layers. Before we get too bogged down, let’s go over the finished product:
The above represents our final neural net with convolutional layers. The whole thing was trainined on your device, and used data from the MNIST database. This is nothing groundbreaking– digit recognition is the traditional neural net example– but it’s a helpful introduction to the topic.