Consistency and fluctuations for stochastic gradient Langevin dynamics
This provides theoretical guarantees and practical tuning recommendations for SGLD, an incremental improvement in scalable Bayesian inference for large-scale data analysis.
The paper tackles the computational inefficiency of standard MCMC algorithms on large datasets by analyzing the stochastic gradient Langevin dynamics (SGLD) method, proving its consistency and deriving an optimal step-size sequence that achieves a mean squared error rate of O(m^{-1/3}).
Applying standard Markov chain Monte Carlo (MCMC) algorithms to large data sets is computationally expensive. Both the calculation of the acceptance probability and the creation of informed proposals usually require an iteration through the whole data set. The recently proposed stochastic gradient Langevin dynamics (SGLD) method circumvents this problem by generating proposals which are only based on a subset of the data, by skipping the accept-reject step and by using decreasing step-sizes sequence $(δ_m)_{m \geq 0}$. %Under appropriate Lyapunov conditions, We provide in this article a rigorous mathematical framework for analysing this algorithm. We prove that, under verifiable assumptions, the algorithm is consistent, satisfies a central limit theorem (CLT) and its asymptotic bias-variance decomposition can be characterized by an explicit functional of the step-sizes sequence $(δ_m)_{m \geq 0}$. We leverage this analysis to give practical recommendations for the notoriously difficult tuning of this algorithm: it is asymptotically optimal to use a step-size sequence of the type $δ_m \asymp m^{-1/3}$, leading to an algorithm whose mean squared error (MSE) decreases at rate $\mathcal{O}(m^{-1/3})$