Machine Learning is Fun! Part 3: Deep Learning and Convolutional Neural Networks

Adam Geitgey

12.2K156

Very nice! Just a few questions:

(1) At “Step 2: Feed each image tile into a small neural network”. Can you please explain a bit on how does the “Small Neural Network” get trained (a CNN with random initial weights)? What does the 5 yellow circle at Outputs mean? Any example?

(2) At “Step 4: Downsampling”, how to choose the max value given that each unit is 5 yellow cirles output? Any example? Thanks.