Autoencoder (AE)


Created: =dateformat(this.file.ctime,"dd MMM yyyy, hh:mm a") | Modified: =dateformat(this.file.mtime,"dd MMM yyyy, hh:mm a") Tags: knowledge


Overview

Introduction

  • auto-encoder is a neural network architecture that encodes an input image onto an embedding layer

Variational Auto-Encoder

  • specialized form of traditional architecture that project the input data onto a probability distribution (usually a Gaussian distribution)
  • 2 loss functions
    • Mean Squared Error (MSE) to calculate loss between output image and GT
    • KL Divergence to calculate statistical distance between true distribution and approximating distribution

Questions

  • how are AEs generative?

Theoretical References

Papers

Articles

Courses


Code References

Methods

Tools, Frameworks