Hamiltonian Monte Carlo Methods in Machine Learning

Markov Chain Monte Carlo (MCMC) methods are considered one of the most influential algorithms for scientific practice in the 21st century. MCMC methods have facilitated the growth in the adoption of principled Bayesian Inference across numerous disciplines. In particular, Hamiltonian Monte Carlo (HMC) methods have revolutionized probabilistic inference in the fields of Machine Learning and Statistics. Hamiltonian Monte Carlo Methods in Machine Learning provides a targeted reference on Hamiltonian Monte Carlo (HMC) methods for practitioners and researchers across numerous application domains. The book offers a comprehensive introduction to Hamiltonian Monte Carlo methods. The book further provides a cutting-edge exposition of the current pathologies of HMC-based methods in both tuning and scaling to sampling complex real-world posteriors. These are mainly in the scaling of inference (e.g., Deep Neural Networks), tuning of performance-sensitive sampling parameters and high sample autocorrelation. The book then traverses numerous solutions to these pitfalls. The authors present the advanced HMC methods with applications in renewable energy, finance and image classification for biomedical applications. Readers of the book will be acquainted with both HMC sampling theory and algorithm implementation. A Python-based code repository of all the algorithms considered is supplied to assist readers with the practical implementation of the algorithms in their work. Hamiltonian Monte Carlo Methods in Machine Learning introduces methods for optimal tuning of HMC parameters, as well as an introduction of Shadow and Non-canonical HMC methods with improvements and speedup. Lastly, the authors address the critical issues of variance reduction for parameter estimates of numerous HMC based samplers.
show more