Disentanglement with beta-VAEs | Deep Learning скачать в хорошем качестве

Disentanglement with beta-VAEs | Deep Learning

Link to my VAE video for a refresher:    • Understanding Variational Autoencoders (VA...   In this video, we explore how and why modifying the VAE loss function enables us to achieve disentanglement in the latent space, with different latent variables corresponding to different semantic features of the data. We take a look at the original beta-VAE formulation, as well as controlled capacity increase, and the total correlation VAE. Timestamps -------------------- 00:00 Introduction 00:35 Variational Autoencoders 02:11 The Importance of Beta 03:15 Understanding Disentangling in beta-VAE 10:41 Controlled Capacity Increase 12:10 Isolating Sources of Disentanglement in VAEs 15:12 beta-TCVAE 15:37 Disentanglement Metrics Papers ----------- Original beta-VAE paper (Higgins et. al, 2016): https://openreview.net/forum?id=Sy2fz... Understanding Disentangling in beta-VAE (Burgess et. al, 2018): https://arxiv.org/abs/1804.03599 Shannon's paper on information theory (Shannon, 1948): https://people.math.harvard.edu/~ctm/... Isolating Sources of Disentanglement in VAEs (Chen et. al, 2019): https://arxiv.org/abs/1802.04942