Reflection, Refraction, and Hamiltonian Monte Carlo

Authors: Hadi Mohasel Afshar, Justin Domke

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that by reducing the number of rejected samples, this method improves on traditional HMC.
Researcher Affiliation Academia Hadi Mohasel Afshar Research School of Computer Science Australian National University Canberra, ACT 0200 hadi.afshar@anu.edu.au Justin Domke National ICT Australia (NICTA) & Australian National University Canberra, ACT 0200 Justin.Domke@nicta.com.au
Pseudocode Yes Algorithm 1: BASELINE & REFLECTIVE HMC ALGORITHMS
Open Source Code No The paper does not provide any explicit statements about the availability of open-source code or links to a code repository.
Open Datasets No The comparison takes place on a heavy tail piecewise model with (non-normalized) negative log probability... (18). This implies a synthetic model/distribution is used, not an existing publicly available dataset.
Dataset Splits No The paper does not specify exact percentages or sample counts for training, validation, or test splits. It describes running Markov chains and evaluating WMAE.
Hardware Specification No All algorithms are implemented in java and run on a single thread of a 3.40GHz CPU. This CPU description is too general and does not provide a specific model number or full hardware specification.
Software Dependencies No All algorithms are implemented in java. No specific version of Java or any other software dependencies with version numbers are mentioned.
Experiment Setup Yes The baseline HMC and RHMC number of steps L and step size ϵ are chosen to be 100 and 0.1 respectively. ... We use a diagonal matrix for A where, for each repetition, each entry on the main diagonal is either exp( 5) or exp(5) with equal probabilities.