Robust and Scalable Bayes via a Median of Subset Posterior Measures
arXiv:1403.2660
Abstract
We propose a novel approach to Bayesian analysis that is provably robust to outliers in the data and often has computational advantages over standard methods. Our technique is based on splitting the data into non-overlapping subgroups, evaluating the posterior distribution given each independent subgroup, and then combining the resulting measures. The main novelty of our approach is the proposed aggregation step, which is based on the evaluation of a median in the space of probability measures equipped with a suitable collection of distances that can be quickly and efficiently evaluated in practice. We present both theoretical and numerical evidence illustrating the improvements achieved by our method.
Cited by in corpus (21)
- Scalable Bayes via Barycenter in Wasserstein Space
- Semi-dual Regularized Optimal Transport
- Informed Proposal Monte Carlo
- A Divide-and-Conquer Bayesian Approach to Large-Scale Kriging
- Expandable Factor Analysis
- Joining and splitting models with Markov melding
- Practical bounds on the error of Bayesian posterior approximations: A nonasymptotic approach
- Consistency and fluctuations for stochastic gradient Langevin dynamics
- Bayesian computation: a perspective on the current state, and sampling backwards and forwards
- A Case Study Competition Among Methods for Analyzing Large Spatial Data
- Bayesian Conditional Density Filtering
- Efficiency Assessment of Approximated Spatial Predictions for Large Datasets
- Scalable and Efficient Statistical Inference with Estimating Functions in the MapReduce Paradigm for Big Data
- Linear-Cost Covariance Functions for Gaussian Random Fields
- A Divide and Conquer Strategy for High Dimensional Bayesian Factor Models
- Nonparametric Bayesian Aggregation for Massive Data
- Communication Efficient Parallel Algorithms for Optimization on Manifolds
- Sequential Markov Chain Monte Carlo for Bayesian Filtering with Massive Data
- Model-free inference on extreme dependence via waiting times
- Bayesian inference in hierarchical models by combining independent posteriors
- An Algorithm for Distributed Bayesian Inference in Generalized Linear Models