Denoising and Variational Autoencoders

This introductory video explains autoencoders for dimensionality reduction, image denoising, reconstruction, and generation. It covers encoder-decoder architectures, latent spaces, reconstruction and latent losses, and training variational autoencoders.

Syllabus

Intro:
Dimensionality reduction
Denoising autoencoders
Variational autoencoders
Training autoencoders
Introduction
Generative models
Variational autoencoders
Dataset of images
Denoising autoencoders
Linear methods
A friendly introduction to deep learning and neural networks
Mapping the real numbers to the interval 0,1
Sigmoid function
Perceptron
Correct noise
Autoencoders as generators
Latent space
Training a neural network - loss function
Training an autoencoder
Training autoencoders
Reconstruction loss Mean squared error
Reconstruction loss log-loss
Training a variational auto encoder