Geoffrey Hinton (who popularized backpropagation in the 80s) recently published his long-awaited paper on "Capsule Networks" - which provide state of the art classification accuracy on the MNIST dataset. In this video I'll go over how Convolutional networks work, some improvements to them over the years, then i'll go over capsule networks in both theory and code.

Code for this video: https://github.com/llSourcell/capsule_networks

Please Subscribe! And like. And comment. That's what keeps me going.

Follow me: Twitter: https://twitter.com/sirajraval Facebook: https://www.facebook.com/sirajology

More learning resources: https://www.youtube.com/watch?v=FTr3n7uBIuE https://www.youtube.com/watch?v=cAICT4Al5Ow https://www.youtube.com/watch?v=qVwm-9P609I https://www.youtube.com/watch?v=LodC7Zm3X8Q https://github.com/naturomics/CapsNet-Tensorflow https://www.reddit.com/r/MachineLearning/comments/78zvc4/r_171009829_dynamic_routing_between_capsules/

Join us in the Wizards Slack channel: http://wizards.herokuapp.com/

And please support me on Patreon: https://www.patreon.com/user?u=3191693 Instagram: https://www.instagram.com/sirajraval/ Instagram: https://www.instagram.com/sirajraval/