This is a Machine Learning project. The GAN and VAE network are built and tested using python3 and Pytorch. I trained them on MNIST handwritten digits images dataset.
This code has been run on the Google cloud server Colaboratory and can generate handwritten digits images successfully.
[1]I.J. Goodfellow, J.Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville,and Y. Bengio, “Generative adversarial networks,” 2014.
[2]D.P. Kingma and M. Welling, “Auto-encoding variational bayes,” 2014.
[3]J.Rocca,Understanding Generative Adversarial Networks (GANs), Jan 7, 2019.https://towardsdatascience.com/understanding-generative-adversarial-networks-gans-cd6e4651a29.
[4]J.Rocca,Understanding Variational Autoencoders (VAEs), Jan 7, 2019.https://towardsdatascience.com/understanding-variational-autoencoders-vaes-f70510919f73.
[5]N.Bertagnolli,Build a Super Simple GAN in PyTorch,Mar 9, 2020.https://towardsdatascience.com/build-a-super-simple-gan-in-pytorch-54ba349920e4.
[6]kevin Frans,Variational Autoencoders Explained, Aug 6, 2016.http://kvfrans.com/variational-autoencoders-explained/.
Some part of the network design is refer on the projects:
[1]https://github.com/lyeoni/pytorch-mnist-GAN
[2]https://github.com/hwalsuklee/tensorflow-mnist-VAE
[3]https://github.com/lyeoni/pytorch-mnist-VAE