Enter the j(r)VAE: divide, (rotate), and order… the cards

By Medium - 2021-03-21

Description

What are joint (rotationally-invariant) variational autoencoders, and why would we need them? The short answer to the first question is that j(r)VAE is a version of a variational autoencoder where one ...

Summary

  • What are joint (rotationally-invariant) variational autoencoders, and why would we need them?
  • Hence, we would like to use unsupervised ML to find relevant traits within the data, and that’s exactly where variational autoencoders with their capability to disentangle data representations and find a priori unknown factors of variability within the data can come in handy.
  • Images by the authors.
  • When we increase the orientation disorder, a = 120, we start to see more than one cluster for each hand, e.g.
  • there is a cluster corresponding to two orientations of clubs and spades rotated by 120 degrees.

 

Topics

  1. Backend (0.35)
  2. Machine_Learning (0.22)
  3. Database (0.16)

Similar Articles

15 Essential Steps To Build Reliable Data Pipelines

By Medium - 2020-12-01

If I learned anything from working as a data engineer, it is that practically any data pipeline fails at some point. Broken connection, broken dependencies, data arriving too late, or some external…