Explore the innovative concept of dynamic routing between capsules in this 31-minute lecture from the University of Central Florida. Delve into the motivation behind this approach, examining its applications in computer graphics and inverse graphics. Learn about capsules and how they differ from traditional neurons, understanding the routing-by-agreement process and the routing algorithm. Discover the CapsNet architecture, including primary capsules and reconstruction techniques. Analyze the loss function and examine results on the MNIST dataset, gaining insights into what capsule dimensions represent. Investigate the performance on multiMNIST and other datasets, broadening your understanding of this cutting-edge neural network architecture.

Syllabus

Motivation
Computer Graphics
Inverse Graphics
Capsules
Traditional Neuron
Routing-by-Agreement
Routing Algorithm
PrimaryCaps
CapsNet Architecture - Reconstruction
Loss Function
Results on MNIST - Reconstruction
What Capsule Dimensions Represent
Results on multiMNIST
Other Dataset