Firefly Monte Carlo: Exact MCMC with Subsets of Data
Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence. Best Papers from Sister Conferences Track. Pages 4289-4294.
Abstract
Markov chain Monte Carlo (MCMC) is a popular tool for Bayesian inference.However, MCMC cannot be practically applied to large data sets because of theprohibitive cost of evaluating every likelihood term at every iteration. Here we present Firefly Monte Carlo (FlyMC) MCMC algorithm with auxiliary variables that only queries the likelihoods of a subset of the data at each iteration yet simulates from the exact posterior distribution. FlyMC is compatible with modern MCMC algorithms, and only requires a lower bound on the per-datum likelihood factors. In experiments, we find that FlyMC generates samples from the posterior more than an order of magnitude faster than regularMCMC, allowing MCMC methods to tackle larger datasets than were previously considered feasible.
@inproceedings{ijcai2015p610,
title = {Firefly Monte Carlo: Exact MCMC with Subsets of Data},
author = {Maclaurin, Dougal and Adams, Ryan P.},
booktitle = {Proceedings of the Twenty-Fourth International Joint Conference on
Artificial Intelligence, {IJCAI-15}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {Qiang Yang and Michael Wooldridge},
pages = {4289--4294},
year = {2015},
month = {7},
note = {Best Papers from Sister Conferences Track},
url = {https://ijcai.org/proceedings/2015/610},
}
TY - CPAPER TI - Firefly Monte Carlo: Exact MCMC with Subsets of Data AU - Maclaurin, Dougal AU - Adams, Ryan P. T2 - Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-15 PB - International Joint Conferences on Artificial Intelligence PY - 2015 SP - 4289 EP - 4294 UR - https://ijcai.org/proceedings/2015/610 ER -Copied
Machine-readable versions of this paper, for tools and AI agents:
