Disentangling by Factorising, Variational Auto-Encoders

PyTorch Implementation of Factorized VAE (Disentangled VAE) Here is a implementation of Factorized VAE based on this paper (MNIST DataSet) Link to the paper https://arxiv.org/pdf/1802.05983.pdf

So, with simple words, it is just VAE, however, with additional regularizer term, which is based on Vanilla GAN, that tries minizmizes the JS distance (not KL, the authors made a mistake) between $q(z)$ and $q(z_1)q(z_2)...q(z_d)$ So, in fact, we have the loss function (ELBO) as the same VAEs (kingma et.al 2014) PLUS JSD divergence of the two terms mentioned above.