Abstract:Bayesian approaches have been successfully integrated into training deep neural networks. One popular family is stochastic gradient Markov chain Monte Carlo methods (SG-MCMC), which have gained increasing interest due to their scalability to handle large datasets and the ability to avoid overfitting. Although standard SG-MCMC methods have shown great performance in a variety of problems, they may be inefficient when the random variables in the target posterior densities have scale differences or are highly correlated. In this work, we present an adaptive Hessian approximated stochastic gradient MCMC method to incorporate local geometric information while sampling from the posterior. The idea is to apply stochastic approximation to sequentially update a preconditioning matrix at each iteration. The preconditioner possesses second-order information and can guide the random walk of a sampler efficiently. Instead of computing and saving the full Hessian of the log posterior, we use limited memory of the sample and their stochastic gradients to approximate the inverse Hessian-vector multiplication in the updating formula. Moreover, by smoothly optimizing the preconditioning matrix, our proposed algorithm can asymptotically converge to the target distribution with a controllable bias under mild conditions. To reduce the training and testing computational burden, we adopt a magnitude-based weight pruning method to enforce the sparsity of the network. Our method is user-friendly and is scalable to standard SG-MCMC updating rules by implementing an additional preconditioner. The sparse approximation of inverse Hessian alleviates storage and computational complexities for large dimensional models. The bias introduced by stochastic approximation is controllable and can be analyzed theoretically. Numerical experiments are performed on several problems.

Stochastic Subgradient MCMC Methods.

Stochastic Gradient Markov Chain Monte Carlo

An adaptive Hessian approximated stochastic gradient MCMC method

Stochastic Gradient MCMC for Nonlinear State Space Models

Laplacian Smoothing Stochastic Gradient Markov Chain Monte Carlo

Fast Sampling Methods for Bayesian Max-Margin Models

CPSG-MCMC: Clustering-Based Preprocessing Method for Stochastic Gradient MCMC.

Stochastic Gradient Geodesic MCMC Methods.

Fast Sampling for Bayesian Max-Margin Models

Stochastic Sub-Sampled Newton Method with Variance Reduction

Impact of Parameter Sparsity on Stochastic Gradient MCMC Methods for Bayesian Deep Learning

Stochastic Gradient Descent as Approximate Bayesian Inference

Stochastic Proximal Gradient Algorithm with Minibatches. Application to Large Scale Learning Models

Scalable MCMC for Mixed Membership Stochastic Blockmodels

Learning Deep Generative Models with Doubly Stochastic MCMC

Particle Optimization in Stochastic Gradient MCMC

Stochastic Block-Coordinate Gradient Projection Algorithms for Submodular Maximization.

Convergence Analysis of Stochastic Gradient Descent with MCMC Estimators

Efficient stochastic optimisation by unadjusted Langevin Monte Carlo. Application to maximum marginal likelihood and empirical Bayesian estimation

Stochastic Gradient Piecewise Deterministic Monte Carlo Samplers

Stochastic Collocation Algorithms Using L(1)-Minimization for Bayesian Solution of Inverse Problems