Bayesian-thinking / Simple-Variational-Autoencoder

A VAE written entirely in Numpy/Cupy

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Variational Autoencoder in NumPy

This is a full implementation of a simple VAE written entirely in Numpy (and Cupy). The code runs very slow on CPU so using a GPU with Cupy is recommended.

Original Paper: Auto-Encoding Variational Bayes, Diederik P Kingma, Max Wellington

Used preprocessing code and inspired by @shinseung428's Vanilla GAN.

Implementation Details

  • Learns the MNIST Dataset
  • Xavier Initialization
  • Adam Optimizer

Requirements

  • Numpy
  • PIL (visualize results)
  • Cupy (Optional)

Network

network

Results

result

Author

K / @kwj2104

About

A VAE written entirely in Numpy/Cupy


Languages

Language:Python 100.0%