On Testing of Samplers

Authors: Kuldeep S Meel, Yash Pralhad Pote, Sourav Chakraborty

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present a prototype implementation of Barbarik2 and use it to test three state-of-the-art samplers. To demonstrate the practical efficiency of Barbarik2, we developed a prototype implementation in Python and performed an experimental evaluation with several samplers.
Researcher Affiliation Academia 1School of Computing, National University of Singapore 2Indian Statistical Institute, Kolkata
Pseudocode Yes Algorithm 1 Barbarik2(G, A, ε, η, δ, ϕ, S, wt) Algorithm 2 Barbarik2Kernel(ϕ, σ1, σ2) Algorithm 3 Bias(ˆσ, Γ, S)
Open Source Code Yes The accompanying tool, available open source, can be found at https://github.com/meelgroup/barbarik.
Open Datasets Yes We conducted our experiments on 72 publicly available benchmarks, which have been employed in the evaluation of samplers proposed in the past [13, 21].
Dataset Splits No The paper uses publicly available benchmarks to evaluate samplers but does not describe any specific training, validation, or test dataset splits for these benchmarks.
Hardware Specification Yes All experiments were conducted on a high performance computing cluster with 600 E5-2690 v3 @2.60GHz CPU cores.
Software Dependencies No The paper mentions 'Python' for implementation and uses 'WAPS' as an ideal sampler, but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup No The paper specifies test parameters like tolerance, intolerance, and confidence values, and mentions instantiating Barbarik2Kernel with m=12 and k=2m-1. However, it does not provide explicit hyperparameters related to model training or system-level training settings as typically found in machine learning experimental setups.