CSC321: Neural Networks

Lecture 13: Learning without a teacher: Autoencoders and Principal Components Analysis

Three problems with backpropagation

Three kinds of learning

The Goals of Unsupervised Learning

Using backprop for unsupervised learning

Self-supervised backprop in a linear network

Principal Components Analysis

A picture of PCA with N=2 and M=1

Self-supervised backprop and clustering

Clustering and backpropagation

A spectrum of representations