On the Exploration of Local Significant Differences For Two-Sample Test

Authors: Zhijian Zhou, Jie Ni, Jia-He Yao, Wei Gao

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We finally conduct extensive experiments to validate the effectiveness of our proposed methods on two-sample test and the exploration of local significant differences.
Researcher Affiliation Academia Zhi-Jian Zhou, Jie Ni, Jia-He Yao, Wei Gao National Key Laboratory for Novel Software Technology, Nanjing University, China School of Artificial Intelligence, Nanjing University, China
Pseudocode Yes Algorithm 1 Construction of partition tree
Open Source Code No The paper mentions downloading code for comparison methods (e.g., 'Dataset blob is downloaded from github.com/fengliu90/DK-for-TST'), but it does not provide an explicit statement or link for the open-source code of its own proposed methods.
Open Datasets Yes Dataset blob is downloaded from github.com/fengliu90/DK-for-TST, and other datasets are downloaded from www.openml.org.
Dataset Splits Yes We train on a subset of each available data, and test on 100 random subsets from the remaining dataset, and the ratio is set as 4 : 1 for training and testing. We repeat such process 10 times for each dataset. ... For our MEMa Bi D, we set α = 0.05 and take 5-fold cross validation to select β [1 : 0.2 : 2].
Hardware Specification Yes All experiments are performed with Python on nodes of a computational cluster with a single CPU (Intel Core i9-10900X 3.7GHz) and a single GPU (Ge Force RTX 2080 Ti), running Ubuntu with 128GB main memory.
Software Dependencies No In optimization, we adapt Adam optimization method from the pytorch library in python [88, 89]. ... running Ubuntu with 128GB main memory. The paper mentions software like Python, PyTorch, Adam, and Ubuntu, but does not provide specific version numbers for these dependencies.
Experiment Setup Yes For our MEMa Bi D, we set α = 0.05 and take 5-fold cross validation to select β [1 : 0.2 : 2]. We limit the cardinality of test locations within 20 for ME and MEMa Bi D as in [9 11], and optimization parameters of Eqn. (5) is presented in Appendix C. ... Table 4 presents the details of the hyperparameter settings for each dataset, including the number of test locations, learning rate.