TzK: Flow-Based Conditional Generative Model

Abstract

We formulate a new class of conditional generative models based on probability flows. Trained with maximum likelihood, it provides efficient inference and sampling from class-conditionals or the joint distribution, and does not require a priori knowledge of the number of classes or the relationships between classes. This allows one to train generative models from multiple, heterogeneous datasets, while retaining strong prior models over subsets of the data (e.g., from a single dataset, class label, or attribute). In this paper, in addition to end-to-end learning, we show how one can learn a single model from multiple datasets with a relatively weak Glow architecture, and then extend it by conditioning on different knowledge types (e.g., a single dataset). This yields log likelihood comparable to state-of-the-art, compelling samples from conditional priors.

8 pages, 9 figures, 2 tables, preprint

References in corpus (14)

• Auto-Encoding Variational Bayes
• NICE: Non-linear Independent Components Estimation
• Variational Inference with Normalizing Flows
• InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
• Density estimation using Real NVP
• MINE: Mutual Information Neural Estimation
• MADE: Masked Autoencoder for Distribution Estimation
• Adversarial Autoencoders
• Realistic Evaluation of Deep Semi-Supervised Learning Algorithms
• Improving Variational Inference with Inverse Autoregressive Flow
• Sylvester Normalizing Flows for Variational Inference
• Learning Latent Subspaces in Variational Autoencoders
• Learning Disentangled Joint Continuous and Discrete Representations
• Implicit Autoencoders