Variational Inference: A Review for Statisticians
By David M. Blei, Alp Kucukelbir, Jon D. McAuliffe. 2016
5,822 citations recorded by the source corpus.
ABSTRACT One of the core problems of modern statistics is to approximate difficult-to-compute probability densities. This problem is especially important in Bayesian statistics, which frames all inference about unknown quantities as a calculation involving the posterior density. In this article, we review variational inference (VI), a method from machine learning that approximates probability densities through optimization. VI has been used in many applications and tends to be faster than classical methods, such as Markov chain Monte Carlo sampling. The idea behind VI is to first posit a family of densities and then to find a member of that family which is close to the target density. Closeness is measured by Kullback–Leibler divergence. We review the ideas behind mean-field variational inference, discuss the special case of VI applied to exponential family models, present a full example with a Bayesian mixture of Gaussians, and derive a variant that uses stochastic optimization to scale up to massive data. We discuss modern research in VI and highlight important open problems. VI is powerful, but it is not yet well understood. Our hope in writing this article is to catalyze statistical research on this class of algorithms. Supplementary materials for this article are available online.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Sergey Levine
- Yoshua Bengio
- Mohit Bansal
- Quoc V. Le
- Jianfeng Gao
- Ruslan Salakhutdinov
- Percy Liang
- Pieter Abbeel
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Ming-Hsuan Yang
- Jitendra Malik
- Xiaogang Wang
- Andrew Zisserman
- Bernt Schiele
In Probabilistic & Bayesian Methods
- David M. Blei
- Qiang Liu
- Stefano Ermon
- Yee Whye Teh
- Michalis K. Titsias
- Max Welling
- Rajesh Ranganath
- Zoubin Ghahramani
Citations in the corpus
- Energy-Inspired Models: Learning with Sampler-Induced Distributions (2019, background): “This allows us to maximize variational lower bounds [32, 7] on the log-likelihood (c.f., Kingma and Welling [37], Rezende et al. [59]).”
- Foundation Posteriors for Approximate Probabilistic Inference (2022, neutral): “Markov Chain Monte Carlo (MCMC) [32, 24] and variational inference (VI) [37, 64, 8] are two widely used examples.”
- Learning World Graphs to Accelerate Hierarchical Reinforcement Learning (2019, positive): “Supposing the state-action trajectories are available, we formulate a recurrent variational inference model (see Section 3.1) [12, 19, 34, 51], treating the action sequences as evidence and, for each state s t in the sequence, inferring a binary latent variable z t that controls whether to keep the…”
- Counterfactual inference for consumer choice across many product categories (2019, methodology): “We instead approximate the posterior distribution using variational inference, which is typically faster on large scale Bayesian problems than classical methods such as Markov Monte Carlo sampling (Blei et al., 2017).”
- Graphite: Iterative Generative Modeling of Graphs (2018, background): “The lower bound is tight when the variational posterior qφ(Z|A,X) matches the true posterior pθ(Z|A,X) and hence maximizing the above objective optimizes for the parameters that define the best approximation to the true posterior within the variational family (Blei et al., 2017).”
- Posterior Meta-Replay for Continual Learning (2021, methodology): “, Blei et al. (2017)) where we look for approximations qθ(W) within a certain family of distributions Q, referred to as the variational family, with members parametrized by θ ∈ Θ.”
- Bayesian Structure Adaptation for Continual Learning (2019, positive): “Therefore, we resort to variational inference (Blei et al., 2017).”
- Uncertainty-guided Continual Learning with Bayesian Neural Networks (2019, methodology): “See (Blei et al., 2017) for a complete review of the theory and (Shridhar et al., 2018) for more discussion on how to use Bayes by Backprop (BBB) in convolutioal neural networks.”
- A deterministic and computable Bernstein-von Mises theorem (2019, methodology): “Finally, the Evidence Lower Bound (ELBO; Blei et al. [2017]) is a well-known lower-bound on the log-integral of a density:”
- Variational Inference of Disentangled Latent Concepts from Unlabeled Observations (2017, background): “…to this, positing a family D of approximate densities over the latents and reducing the approximate inference problem to finding a member density that minimizes the Kullback-Leibler divergence to the true posterior, i.e., q ∗ x = min q ∈D KL( q ( z ) (cid:107) p θ ( z | x )) (Blei et al., 2017).”
- On Variational Bounds of Mutual Information (2019, methodology): “To overcome these scaling difficulties, recent work combines variational bounds (Blei et al., 2017; Donsker & Varad-han, 1983; Barber & Agakov, 2003; Nguyen et al., 2010; Foster et al., 2018) with deep learning (Alemi et al., 2016; 2017; van den Oord et al., 2018; Hjelm et al., 2018; Belghazi et…”
- Normalizing Flows for Probabilistic Modeling and Inference (2019, methodology): “The reverse KL divergence is often used for variational inference (Wainwright and Jordan, 2008; Blei et al., 2017), a form of approximate Bayesian inference.”
- VAE with a VampPrior (2017, background): “In recent years, different approaches to achieving this goal were proposed by formulating alternative training objectives to the loglikelihood [11, 13, 23] or by utilizing variational inference [3].”
- Conditioning by adaptive sampling for robust design (2019, methodology): “Although one might consider using an Evidence Lower Bound (ELBO) (Blei et al., 2017) approximation”