2020/04/11 by Xuezhe Ma, Ma, Xuezhe, Xiang Kong +5 · 1 citation
Computer Science · Engineering · Mathematics · #Algorithm #Algorithms and Data Compression #Artificial Intelligence in Games #Artificial intelligence #Computer Graphics and Visualization Techniques #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Control engineering #Control theory (sociology) #Decoupling (probability) #Embedding #Encoding (memory) #Engineering #FOS: Computer and information sciences #FOS: Electrical engineering #Feature learning #Generative Adversarial Networks and Image Synthesis #Generative grammar #Generative model #Image and Video Processing (eess.IV) #Invertible matrix #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine learning #Mathematics #Music and Audio Processing #Pattern recognition (psychology) #Pure mathematics #Representation (politics) #Theoretical computer science #Unsupervised learning #cs.CV #cs.LG #eess.IV #electronic engineering #information engineering #stat.ML
paper · pdf · doi:10.48550/arxiv.2004.11820
published in arXiv (Cornell University) (Cornell University) · Camera-ready at ICLR 2021. 23 pages (plus appendix), 16 figures, 5 tables. Due to arxiv size constraints, this version is using downscaled images. Please download the full-resolution version from https://vixra.org/abs/2004.0222
openalex publication_date 2020/04/12 · arxiv created 2021/03/15 · arxiv updated 2021/03/17 · openalex created_date 2021/03/29 · openalex updated_date 2026/08/05
In this work, we propose a new generative model that is capable of automatically decoupling global and local representations of images in an entirely unsupervised setting, by embedding a generative flow in the VAE framework to model the decoder. Specifically, the proposed model utilizes the variational auto-encoding framework to learn a (low-dimensional) vector of latent variables to capture the global information of an image, which is fed as a conditional input to a flow-based invertible decoder with architecture borrowed from style transfer literature. Experimental results on standard image benchmarks demonstrate the effectiveness of our model in terms of density estimation, image generation and unsupervised representation learning. Importantly, this work demonstrates that with only architectural inductive biases, a generative model with a likelihood-based objective is capable of learning decoupled representations, requiring no explicit supervision. The code for our model is available at https://github.com/XuezheMax/wolf.