Anytime Sampling for Autoregressive Models via Ordered Autoencoding
arXiv:2102.11495
Abstract
Autoregressive models are widely used for tasks such as image and audio generation. The sampling process of these models, however, does not allow interruptions and cannot adapt to real-time computational resources. This challenge impedes the deployment of powerful autoregressive models, which involve a slow sampling process that is sequential in nature and typically scales linearly with respect to the data dimension. To address this difficulty, we propose a new family of autoregressive models that enables anytime sampling. Inspired by Principal Component Analysis, we learn a structured representation space where dimensions are ordered based on their importance with respect to reconstruction. Using an autoregressive model in this latent space, we trade off sample quality for computational efficiency by truncating the generation process before decoding into the original data space. Experimentally, we demonstrate in several image and audio generation tasks that sample quality degrades gracefully as we reduce the computational budget for sampling. The approach suffers almost no loss in sample quality (measured by FID) using only 60\% to 80\% of all latent dimensions for image data. Code is available at https://github.com/Newbeeer/Anytime-Auto-Regressive-Model .
Accepted by ICLR 2021
References in corpus (10)
- Auto-Encoding Variational Bayes
- Denoising Diffusion Probabilistic Models
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- Neural Discrete Representation Learning
- Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
- Generative Modeling by Estimating Gradients of the Data Distribution
- MADE: Masked Autoencoder for Distribution Estimation
- Neural Machine Translation in Linear Time
- Adversarial Audio Synthesis
- Learning Ordered Representations with Nested Dropout