Spherical Hamiltonian Monte Carlo for Constrained Target Distributions

Authors: Shiwei Lan, Bo Zhou, Babak Shahbaba

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we evaluate our proposed methods, Spherical HMC, by comparing its efficiency to that of Random Walk Metropolis (RWM) and Wall HMC using simulated and real data. To this end, we define efficiency in terms of time-normalized effective sample size (ESS). Given B MCMC samples for each parameter, ESS = B[1+ 2ΣK k=1γ(k)] 1, where ΣK k=1γ(k) is the sum of K monotone sample autocorrelations (Geyer, 1992). We use the minimum ESS normalized by the CPU time, s (in seconds), as the overall measure of efficiency: min(ESS)/s.
Researcher Affiliation Academia Shiwei Lan SLAN@UCI.EDU Department of Statistics, University of California, Irvine, CA 92697, USA. Bo Zhou BZHOU1@UCI.EDU Department of Statistics, University of California, Irvine, CA 92697, USA. Babak Shahbaba BABAKS@UCI.EDU Department of Statistics, University of California, Irvine, CA 92697, USA.
Pseudocode Yes Algorithm 1 Spherical HMC
Open Source Code Yes All computer codes are available online at http://www.ics.uci.edu/ slan/Sph HMC.
Open Datasets Yes We evaluate our method based on the diabetes data set (N=442, D=10) discussed in (Park & Casella, 2008).
Dataset Splits No The paper mentions obtaining MCMC samples and discarding initial ones but does not specify training, validation, or test dataset splits or cross-validation methods.
Hardware Specification No The paper does not provide specific hardware details such as CPU/GPU models, memory, or cloud computing resources used for experiments.
Software Dependencies No The paper does not provide specific software names with version numbers, nor does it list any versioned libraries or solvers used for its implementation or experiments.
Experiment Setup Yes For Wall HMC and Spherical HMC, we fix the number of leapfrog steps to 10 and set the trajectory length such that they both have comparable acceptance rates around 70%.