Austerity in MCMC Land: Cutting the Metropolis-Hastings Budget

Authors: Anoop Korattikara, Yutian Chen, Max Welling

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In the next section we introduce the MH algorithm and discuss its drawbacks. Then in Section 3, we introduce the idea of approximate MCMC methods and the bias variance trade-off involved. We develop approximate MH tests for Bayesian posterior sampling in Section 4 and present a theoretical analysis in Section 5. Finally, we show our experimental results in Section 6 and conclude in Section 7.
Researcher Affiliation Academia Anoop Korattikara AKORATTI@UCI.EDU School of Information & Computer Sciences, University of California, Irvine, CA 92617, USA Yutian Chen YUTIAN.CHEN@ENG.CAM.EDU Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, UK Max Welling WELLING@ICS.UCI.EDU Informatics Institute, University of Amsterdam, Science Park 904 1098 XH, Amsterdam, Netherlands
Pseudocode Yes Algorithm 1 Approximate MH test
Open Source Code No The paper does not provide any explicit statement or link to open-source code for the methodology described.
Open Datasets Yes The target distribution in this experiment was the posterior for a logistic regression model trained on the MNIST dataset for classifying digits 7 vs 9. The dataset consisted of 12214 datapoints and we reduced the dimensionality from 784 to 50 using PCA. ... We applied this to the Mini Boo NE dataset from the UCI machine learning repository(Bache & Lichman, 2013).
Dataset Splits Yes We randomly split the data into a training (80%) and testing (20%) set.
Hardware Specification No The paper does not specify any hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers.
Experiment Setup Yes We chose a zero mean spherical Gaussian prior with precision = 10, and set σRW = 0.01. ... We set λ = 10 10 in this experiment. ... The acceptance rate for update moves is kept at 50%.