Change point detection and inference in multivariate non-parametric models under mixing conditions
Authors: Carlos Misael Madrid Padilla, Haotian Xu, Daren Wang, OSCAR HERNAN MADRID PADILLA, Yi Yu
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical studies are provided to complement our theoretical findings. and 5 Numerical Experiments |
| Researcher Affiliation | Academia | Carlos Misael Madrid Padilla Department of Mathematics University of Notre Dame cmadridp@nd.edu Haotian Xu Department of Statistics University of Warwick haotian.xu.1@warwick.ac.uk Daren Wang Department of Statistics University of Notre Dame dwang24@nd.edu Oscar Hernan Madrid Padilla Department of Statistics University of California, Los Angeles oscar.madrid@stat.ucla.edu Yi Yu Department of Statistics University of Warwick yi.yu.2@warwick.ac.uk |
| Pseudocode | Yes | Algorithm 1 Multivariate non-parametric Seeded Binary Segmentation. and Algorithm 2 Long-run variance estimators |
| Open Source Code | No | The code used for numerical experiments is available upon request prior to publication. |
| Open Datasets | Yes | The stock price data are downloaded from https://fred.stlouisfed.org/series. |
| Dataset Splits | No | The paper does not explicitly provide specific percentages or sample counts for training, validation, or test dataset splits. It describes simulation settings and repetitions of experiments, but not data partitioning for model evaluation. |
| Hardware Specification | Yes | The tests were conducted on a machine powered by an Apple M2 chip with an 8-core CPU. |
| Software Dependencies | Yes | using corresponding R functions in changepoints (Xu, Padilla, Wang & Li 2022), ecp (James et al. 2019) and hdbinseg (Cho & Fryzlewicz 2018) packages. and The Subregion-Adaptive Vegas Algorithm is available in R package cubature (Narasimhan et al. 2022) |
| Experiment Setup | Yes | Preliminary estimators are set as h = 2 (1/T)1/(2r+p), while the second stage estimator has bandwidths respectively set as eh = 0.05 and h1 = 2 bκ1/r k . Selection of R = max b K k=1{ek sk} 3/5 with {(sk, ek)} b K k=1 is guided by Theorem 3 using {(sk, ek)} b K k=1 from (4). |