Towards Conceptual Compression
arXiv:1604.08772
Abstract
We introduce a simple recurrent variational auto-encoder architecture that significantly improves image modeling. The system represents the state-of-the-art in latent variable models for both the ImageNet and Omniglot datasets. We show that it naturally separates global conceptual information from lower level details, thus addressing one of the fundamentally desired properties of unsupervised learning. Furthermore, the possibility of restricting ourselves to storing only global information about an image allows us to achieve high quality 'conceptual compression'.
14 pages, 13 figures
Cited by in corpus (25)
- An Introduction to Variational Autoencoders
- Variational image compression with a scale hyperprior
- End-to-end Optimized Image Compression
- Image and Video Compression with Neural Networks: A Review
- Lossy Image Compression with Compressive Autoencoders
- Improving Variational Inference with Inverse Autoregressive Flow
- Dynamical Variational Autoencoders: A Comprehensive Review
- Learning for Video Compression
- Scalable Population Synthesis with Deep Generative Modeling
- Improved Variational Autoencoders for Text Modeling using Dilated Convolutions
- Cognitive Psychology for Deep Neural Networks: A Shape Bias Case Study
- Comparison of Maximum Likelihood and GAN-based training of Real NVPs
- Simple and Effective VAE Training with Calibrated Decoders
- Semi-Amortized Variational Autoencoders
- Neural Canonical Transformation with Symplectic Flows
- Tighter Variational Bounds are Not Necessarily Better
- Improved Lossy Image Compression with Priming and Spatially Adaptive Bit Rates for Recurrent Networks
- Deep Generative Models for Distribution-Preserving Lossy Compression
- Associative Compression Networks for Representation Learning
- PixelCNN Models with Auxiliary Variables for Natural Image Modeling
- Probing Emergent Semantics in Predictive Agents via Question Answering
- VQ-DRAW: A Sequential Discrete VAE
- Channel-Recurrent Autoencoding for Image Modeling
- Variational Bayesian Quantization
- DeepSIC: Deep Semantic Image Compression