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.

Precomputing strategy for Hamiltonian Monte Carlo method based on regularity in parameter space

A Hybrid Stochastic Gradient Hamiltonian Monte Carlo Method

Hamiltonian Monte Carlo Acceleration Using Surrogate Functions with Random Bases

Efficient History Matching Using the Markov-Chain Monte Carlo Method by Means of the Transformed Adaptive Stochastic Collocation Method

Multi-fidelity Hamiltonian Monte Carlo

Probabilistic Path Hamiltonian Monte Carlo

Quasi-Newton Hamiltonian Monte Carlo.

Variational Hybrid Monte Carlo for Efficient Multi-Modal Data Sampling

An adaptive Hessian approximated stochastic gradient MCMC method

Semi-Separable Hamiltonian Monte Carlo for Inference in Bayesian Hierarchical Models

Repelling-Attracting Hamiltonian Monte Carlo

Accelerating Multilevel Markov Chain Monte Carlo Using Machine Learning Models

Neural Network Gradient Hamiltonian Monte Carlo

Sampling from high-dimensional, multimodal distributions using automatically tuned, tempered Hamiltonian Monte Carlo

Overcoming Ergodicity Problems of the Hybrid Monte Carlo Method using Radial Updates

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

Lagrangian Dynamical Monte Carlo

NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport

Stochastic Gradient Markov Chain Monte Carlo

Efficient MCMC for Gibbs Random Fields using pre-computation

Efficient Sampling on Riemannian Manifolds via Langevin MCMC