Variational Bayesian Inference with Stochastic Search

Abstract

Mean-field variational inference is a method for approximate Bayesian posterior inference. It approximates a full posterior distribution with a factorized set of distributions by maximizing a lower bound on the marginal likelihood. This requires the ability to integrate a sum of terms in the log joint likelihood using this factorized distribution. Often not all integrals are in closed form, which is typically handled by using a lower bound. We present an alternative algorithm based on stochastic optimization that allows for direct optimization of the variational lower bound. This method uses control variates to reduce the variance of the stochastic search gradient, in which existing lower bounds can play an important role. We demonstrate the approach on two non-conjugate models: logistic regression and an approximation to the HDP.

Appears in Proceedings of the 29th International Conference on Machine Learning (ICML 2012)

Cited by in corpus (101)

• Auto-Encoding Variational Bayes
• Variational Inference: A Review for Statisticians
• An Introduction to Variational Autoencoders
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Bayesian Convolutional Neural Networks with Bernoulli Approximate Variational Inference
• Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems
• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• Deep and Confident Prediction for Time Series at Uber
• Neural Variational Inference and Learning in Belief Networks
• Improving Variational Inference with Inverse Autoregressive Flow
• BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling
• Black box variational inference for state space models
• Deep AutoRegressive Networks
• Q-Prop: Sample-Efficient Policy Gradient with An Off-Policy Critic
• Deep Probabilistic Programming
• Fixed-Form Variational Posterior Approximation through Stochastic Linear Regression
• Monte Carlo Gradient Estimation in Machine Learning
• WHAI: Weibull Hybrid Autoencoding Inference for Deep Topic Modeling
• Pathwise Derivatives Beyond the Reparameterization Trick
• Variational Bayes with Synthetic Likelihood
• Generalized Variational Inference: Three arguments for deriving new Posteriors
• Dropout as a Bayesian Approximation: Appendix
• Nonlinear Kalman Filtering with Divergence Minimization
• Structured Stochastic Variational Inference
• Sticking to the Facts: Confident Decoding for Faithful Data-to-Text Generation
• Statistical Decision Making for Optimal Budget Allocation in Crowd Labeling
• The price of debiasing automatic metrics in natural language evaluation
• Black Box Variational Inference
• Discrete Variational Autoencoders
• A physics-aware, probabilistic machine learning framework for coarse-graining high-dimensional systems in the Small Data regime
• Stochastic gradient variational Bayes for gamma approximating distributions
• MuProp: Unbiased Backpropagation for Stochastic Neural Networks
• Bayesian model and dimension reduction for uncertainty propagation: applications in random media
• Towards Safe Machine Learning for CPS: Infer Uncertainty from Training Data
• A probabilistic generative model for semi-supervised training of coarse-grained surrogates and enforcing physical constraints through virtual observables
• Stochastic Variational Inference
• Advances in Variational Inference
• Deep Ensemble Bayesian Active Learning : Addressing the Mode Collapse issue in Monte Carlo dropout via Ensembles
• Approximate Inference with Amortised MCMC
• Semi-Implicit Variational Inference
• A practical tutorial on Variational Bayes
• Semi-Conditional Normalizing Flows for Semi-Supervised Learning
• Learning Wake-Sleep Recurrent Attention Models
• Stochastic variational inference for large-scale discrete choice models using adaptive batch sizes
• Online Learning to Sample
• Deeper Connections between Neural Networks and Gaussian Processes Speed-up Active Learning
• Gaussian variational approximation for high-dimensional state space models
• Estimating Gradients for Discrete Random Variables by Sampling without Replacement
• Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data
• Spatial Uncertainty Sampling for End-to-End Control
• Reparameterization Gradients through Acceptance-Rejection Sampling Algorithms
• Overdispersed Black-Box Variational Inference
• Determinantal Point Processes for Mini-Batch Diversification
• f-Divergence Variational Inference
• Markovian Score Climbing: Variational Inference with KL(p||q)
• Spike and slab variational Bayes for high dimensional logistic regression
• Big Learning with Bayesian Methods
• A stochastic variational framework for fitting and diagnosing generalized linear mixed models
• Infinite-dimensional gradient-based descent for alpha-divergence minimisation
• Learning document embeddings along with their uncertainties
• Efficient Marginalization of Discrete and Structured Latent Variables via Sparsity
• Reparameterization Gradient for Non-differentiable Models
• DisARM: An Antithetic Gradient Estimator for Binary Latent Variables
• Using Large Ensembles of Control Variates for Variational Inference
• Local Expectation Gradients for Doubly Stochastic Variational Inference
• Order Matters: Probabilistic Modeling of Node Sequence for Graph Generation
• On importance-weighted autoencoders
• Causal Mediation Analysis Leveraging Multiple Types of Summary Statistics Data
• Backprop-Q: Generalized Backpropagation for Stochastic Computation Graphs
• Towards a Kernel based Uncertainty Decomposition Framework for Data and Models
• Variance Reduction for Evolution Strategies via Structured Control Variates
• Physics-constrained, data-driven discovery of coarse-grained dynamics
• Variational Approximation of Factor Stochastic Volatility Models
• Adaptive Correlated Monte Carlo for Contextual Categorical Sequence Generation
• Deep Autoencoding Topic Model with Scalable Hybrid Bayesian Inference
• A PAC-Bayesian Perspective on Structured Prediction with Implicit Loss Embeddings
• Learning a Representation Map for Robot Navigation using Deep Variational Autoencoder
• Pathwise Derivatives for Multivariate Distributions
• Bayes EMbedding (BEM): Refining Representation by Integrating Knowledge Graphs and Behavior-specific Networks
• Variational Inference for Gaussian Process with Panel Count Data
• A Rule for Gradient Estimator Selection, with an Application to Variational Inference
• Amortized variance reduction for doubly stochastic objectives
• A Quaternion-Valued Variational Autoencoder
• Double Control Variates for Gradient Estimation in Discrete Latent Variable Models
• Incremental Variational Inference for Latent Dirichlet Allocation
• CARMS: Categorical-Antithetic-REINFORCE Multi-Sample Gradient Estimator
• Quasi-Newton Quasi-Monte Carlo for variational Bayes
• Generalized Transformation-based Gradient
• Stochastic Variational Inference via Upper Bound
• A Divergence Bound for Hybrids of MCMC and Variational Inference and an Application to Langevin Dynamics and SGVI
• Manifold Optimization Assisted Gaussian Variational Approximation
• Multilevel Monte Carlo Variational Inference
• Online Label Aggregation: A Variational Bayesian Approach
• Towards Verified Stochastic Variational Inference for Probabilistic Programs
• Mixed Variational Inference
• Information Theoretic Structured Generative Modeling
• Variational Wasserstein Barycenters with c-Cyclical Monotonicity
• Online Inference for Relation Extraction with a Reduced Feature Set
• An adaptive mixture-population Monte Carlo method for likelihood-free inference
• Approximation Based Variance Reduction for Reparameterization Gradients
• A Collective Variational Autoencoder for Top-$N$ Recommendation with Side Information