Stochastic Gradient Monomial Gamma Sampler

Authors: Yizhe Zhang, Changyou Chen, Zhe Gan, Ricardo Henao, Lawrence Carin

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental It is shown that the proposed approach is better at exploring complex multimodal posterior distributions, as demonstrated on multiple applications and in comparison with other stochastic gradient MCMC methods.
Researcher Affiliation Academia 1Duke University, Durham, NC, 27708. Correspondence to: Yizhe Zhang <yizhe.zhang@duke.edu>.
Pseudocode Yes The complete update scheme, with Euler integrator, for SGMGT is presented in the SM.
Open Source Code No The paper does not provide an explicit statement or link for open-source code for the described methodology.
Open Datasets Yes We evaluated the mixing efficiency and accuracy of SGMGT and SGMGT-D using Bayesian logistic regression (BLR) on 6 real-world datasets from the UCI repository (Bache & Lichman, 2013): German credit (G), Australian credit (A), Pima Indian (P), Heart (H), Ripley (R) and Caravan (C).
Dataset Splits Yes We use 80% of the documents for training and the remaining 20% for testing.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments.
Software Dependencies No The paper mentions models and algorithms but does not specify version numbers for any software dependencies or libraries used for implementation (e.g., PyTorch, TensorFlow, scikit-learn versions).
Experiment Setup Yes We set the minibatch size to 16. Other hyperparameters are provided in the SM. For each experiment, we draw 5000 iterations with 1000 burn-in samples.