Variational AutoEncoders (VAE) with PyTorch - Alexander Van de Kleut
Imagine that we have a large, high-dimensional dataset. For example, imagine we have a dataset consisting of thousands of images. Each image is made up of hundreds of pixels, so each data point has hundreds of dimensions. The manifold hypothesis states that real-world high-dimensional data actually consists of low-dimensional data that is embedded in the high-dimensional space. This means that, while the actual data itself might have hundreds of dimensions, the underlying structure of the data can be sufficiently described using only a few dimensions. This is the motivation behind dimensionality reduction techniques, which try to take high-dimensional data and project it onto a lower-dimensional surface. For humans who visualize most things in 2D (or sometimes 3D), this usually means projecting the data onto a 2D surface. Examples of dimensionality reduction techniques include principal component analysis (PCA) and t-SNE. Chris Olah’s blog has a great post reviewing some dimensionality
Download the jupyter notebook and run this blog post yourself! Motivation Imagine that we have a large, high-dimensional dataset. For example, imagine we have a dataset consisting of thousands of images. Each image is made up of hundreds of pixels, so each data point has hundreds of dimensions. The manifold hypothesis states that real-world high-dimensional data actually consists of low-dimensional data that is embedded in the high-dimensional space. This means that, while the actual data itself might have hundreds of dimensions, the underlying structure of the data can be sufficiently describ
related reading
- Difference between AutoEncoder (AE) and Variational AutoEncoder (VAE) | Towards Data Sciencetowardsdatascience.com
- Variational autoencoders.jeremyjordan.me
- Tutorial - What is a variational autoencoder? – Jaan Lı 李jaan.io
- Tutorial on Variational Autoencodersarxiv.org
- Understanding VQ-VAE (DALL-E Explained Pt. 1)mlberkeley.substack.com
- Jupyter Notebook Viewernbviewer.org
- Generative modelling in latent space – Sander Dielemansander.ai
- From Autoencoder to Beta-VAE | Lil'Loglilianweng.github.io
- Variational autoencoder - Wikipediaen.wikipedia.org
- [1904.01277] A PCA-like Autoencoderarxiv.org
- [1906.02691] An Introduction to Variational Autoencodersarxiv.org
- jotterbach.github.io/content/posts/autoencoders/2016-07-18-AutoEncoders/jotterbach.github.io