• Corpus ID: 198968035

Variational f-divergence Minimization

  title={Variational f-divergence Minimization},
  author={Mingtian Zhang and Thomas Bird and Raza Habib and Tianlin Xu and David Barber},
Probabilistic models are often trained by maximum likelihood, which corresponds to minimizing a specific f-divergence between the model and data distribution. In light of recent successes in training Generative Adversarial Networks, alternative non-likelihood training criteria have been proposed. Whilst not necessarily statistically efficient, these alternatives may better match user requirements such as sharp image generation. A general variational method for training probabilistic latent… 

Figures and Tables from this paper

Spread Divergences

This work defines a spread divergence on modified p and q and describes sufficient conditions for the existence of such a divergence and demonstrates how to maximize the discriminatory power of a given divergence by parameterizing and learning the spread.

Imitation Learning as f-Divergence Minimization

This work proposes a general imitation learning framework for estimating and minimizing any f-Divergence, and shows that the approximate I-projection technique is able to imitate multi-modal behaviors more reliably than GAIL and behavior cloning.

Posterior-Aided Regularization for Likelihood-Free Inference

This paper proposes a universally applicable regularization technique, called Posterior-Aided Regularization (PAR), which is applicable to learning the density estimator, regardless of the model structure, and theoretically proves the asymptotic convergence of the regularized optimal solution to the unregularized optimal solutions as the regularization magnitude converges to zero.

Low-Discrepancy Points via Energetic Variational Inference

A deterministic variational inference approach and generate lowdiscrepancy points by minimizing the kernel discrepancy, also known as the Maximum Mean Discrepancy or MMD, is proposed and the resulting algorithm is named EVI-MMD.

f-Divergence Variational Inference

The $f$-VI framework not only unifies a number of existing VI methods, but offers a standardized toolkit for VI subject to arbitrary divergences from the f-divergence family, and provides a sandwich estimate of marginal likelihood (or evidence).

GFlowNets and variational inference

This paper builds bridges between two families of probabilistic algorithms: (hi-erarchical) variational inference (VI), which is typically used to model distributions over continuous spaces, and

α-VAEs : Optimising variational inference by learning data-dependent divergence skew

The skew-geometric Jensen-Shannon divergence ( JSα ) allows for an intuitive interpolation between forward and reverse Kullback-Leibler (KL) divergence based on the skew parameter α. While the

Neural Posterior Regularization for Likelihood-Free Inference

This paper introduces a regularization technique, namely Neural Posterior Regularization (NPR), which enforces the model to explore the input parameter space e ff ectively and empirically validate that NPR attains the statistically significant gain on benchmark performances for diverse simulation tasks.

Constraining Variational Inference with Geometric Jensen-Shannon Divergence

This work presents a regularisation mechanism based on the skew geometric-Jensen-Shannon divergence, motivated by limiting cases, which leads to an intuitive interpolation between forward and reverse KL in the space of both distributions and divergences.



f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization

It is shown that any f-divergence can be used for training generative neural samplers and the benefits of various choices of divergence functions on training complexity and the quality of the obtained generative models are discussed.

Variational Inference using Implicit Distributions

This paper provides a unifying review of existing algorithms establishing connections between variational autoencoders, adversarially learned inference, operator VI, GAN-based image reconstruction, and more, and provides a framework for building new algorithms.

Stabilizing Training of Generative Adversarial Networks through Regularization

This work proposes a new regularization approach with low computational cost that yields a stable GAN training procedure and demonstrates the effectiveness of this regularizer accross several architectures trained on common benchmark image generation tasks.

Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks

Adversarial Variational Bayes (AVB), a technique for training Variational Autoencoders with arbitrarily expressive inference models by introducing an auxiliary discriminative network that allows to rephrase the maximum-likelihood-problem as a two-player game, hence establishing a principled connection between VAEs and Generative Adversarial Networks (GANs).

Adversarial Feature Learning

Bidirectional Generative Adversarial Networks are proposed as a means of learning the inverse mapping of GANs, and it is demonstrated that the resulting learned feature representation is useful for auxiliary supervised discrimination tasks, competitive with contemporary approaches to unsupervised and self-supervised feature learning.

Auto-Encoding Variational Bayes

A stochastic variational inference and learning algorithm that scales to large datasets and, under some mild differentiability conditions, even works in the intractable case is introduced.

Learning in Implicit Generative Models

This work develops likelihood-free inference methods and highlight hypothesis testing as a principle for learning in implicit generative models, using which it is able to derive the objective function used by GANs, and many other related objectives.

Adversarially Learned Inference

The adversarially learned inference (ALI) model is introduced, which jointly learns a generation network and an inference network using an adversarial process and the usefulness of the learned representations is confirmed by obtaining a performance competitive with state-of-the-art on the semi-supervised SVHN and CIFAR10 tasks.

Spread Divergences

This work defines a spread divergence on modified p and q and describes sufficient conditions for the existence of such a divergence and demonstrates how to maximize the discriminatory power of a given divergence by parameterizing and learning the spread.

Rényi Divergence Variational Inference

The variational R\'enyi bound (VR) is introduced that extends traditional variational inference to R‐enyi's alpha-divergences, and a novel variational inferred method is proposed as a new special case in the proposed framework.