The Randomized Midpoint Method for Log-Concave Sampling

Authors: Ruoqi Shen, Yin Tat Lee

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Numerical Experiments In this section, we compare the algorithm from our paper, randomized midpoint method, with the one from [10]. We test the algorithms on the liver-disorders dataset and the breast-cancer dataset from UCL machine learning [17]. In both datasets, we observe a set of independent samples {xi, yi}m i=1, where yi is the label, xi is the feature and m is the number of samples. We sample from the target distribution p (θ) exp ( f(θ)) , where P m i=1 log exp yix T i θ + 1 , for regularization parameters λ. We set λ to be 10 2 in our experiments. Figure 1 shows the error of randomized midpoint method and the algorithm from [10] with different step size h.
Researcher Affiliation Collaboration Ruoqi Shen University of Washington shenr3@cs.washington.edu Yin Tat Lee University of Washington and Microsoft Research yintat@uw.edu
Pseudocode Yes Algorithm 1 Randomized Midpoint Method for ULD
Open Source Code No The paper does not provide an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We test the algorithms on the liver-disorders dataset and the breast-cancer dataset from UCL machine learning [17].
Dataset Splits No The paper mentions using datasets for numerical experiments but does not provide specific details on training, validation, or test splits (e.g., percentages, sample counts, or explicit splitting methodology).
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper describes numerical experiments but does not specify any software dependencies (e.g., libraries, frameworks, or programming languages) with their corresponding version numbers.
Experiment Setup Yes We set λ to be 10 2 in our experiments.