Stochastic Variational Inference for Hidden Markov Models
This work addresses computational challenges in Bayesian analysis for time-dependent data, such as in genomics, though it is incremental as it extends existing methods to a new setting.
The paper tackled the problem of applying stochastic variational inference to hidden Markov models by developing an algorithm that addresses dependencies in time-dependent data, demonstrating effectiveness on synthetic experiments and a large genomics dataset where batch methods were infeasible.
Variational inference algorithms have proven successful for Bayesian analysis in large data settings, with recent advances using stochastic variational inference (SVI). However, such methods have largely been studied in independent or exchangeable data settings. We develop an SVI algorithm to learn the parameters of hidden Markov models (HMMs) in a time-dependent data setting. The challenge in applying stochastic optimization in this setting arises from dependencies in the chain, which must be broken to consider minibatches of observations. We propose an algorithm that harnesses the memory decay of the chain to adaptively bound errors arising from edge effects. We demonstrate the effectiveness of our algorithm on synthetic experiments and a large genomics dataset where a batch algorithm is computationally infeasible.