Auxiliary Variational MCMC

Authors: Raza Habib, David Barber

ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our sampler on a number of challenging distributions, where the underlying structure is known, and on the task of posterior sampling in Bayesian logistic regression. Code to reproduce all experiments is available at https://github.com/AVMCMC. ... 3 EXPERIMENTS
Researcher Affiliation Academia Raza. Habib Department of Computer Science University College London raza.habib@cs.ucl.ac.uk David. Barber Department of Computer Science University College London david.barber@ucl.ac.uk
Pseudocode Yes Algorithm 1: Auxiliary variational sampler
Open Source Code Yes Code to reproduce all experiments is available at https://github.com/AVMCMC.
Open Datasets Yes We use the heart data-set used by Song et al. (2017), which has 13 covariates and 270 data-points.
Dataset Splits No The paper mentions using the 'heart data-set used by Song et al. (2017)' but does not specify any explicit training, validation, or test dataset splits (e.g., percentages, counts, or methods for creating splits).
Hardware Specification Yes All methods were implemented using Tensorflow 1.10 (Abadi & Agarwal, 2015) and run on a single Tesla K80 GPU.
Software Dependencies Yes All methods were implemented using Tensorflow 1.10 (Abadi & Agarwal, 2015)
Experiment Setup Yes For HMC we use a number of initial runs to select an appropriate step-size and then tune the number of leapfrog steps. ... For the mixtures of Gaussians and ring density experiments we used an auxiliary dimension of 1 and the target dimension was 2. ... For Bayesian logistic regression, the target dimension was 14, the auxiliary dimension used was 2 dimensional and the number of hidden units was 300. The structure was otherwise the same as for the low dimensional experiments.