Orderly Subspace Clustering

Authors: Jing Wang, Atsushi Suzuki, Linchuan Xu, Feng Tian, Liang Yang, Kenji Yamanishi5264-5272

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results with several benchmarks have demonstrated that aside from more accurate clustering against state-of-the-arts, OSC interprets orderly data structure which is beyond what current approaches can offer.
Researcher Affiliation Academia 1Graduate School of Information Science and Technology, The University of Tokyo, Japan 2Faculty of Science and Technology, Bournemouth University, UK 3School of Artificial Intelligence, Hebei University of Technology, China
Pseudocode Yes The whole procedure of OSC is summarized in Algorithm 1. Algorithm 1 Solving OSC
Open Source Code No The paper does not contain any statement about releasing source code for the described methodology or a link to a code repository.
Open Datasets Yes To demonstrate the effectiveness of OSC, we conducted experiments on a variety of datasets, including image datasets (Extended-Yale B, USPS), a document dataset (20newsgroups) and a motion sequence (Hopkins 155).
Dataset Splits No The paper describes how supervision data was selected ('randomly selected 10% data for each dataset, and then constructed 30 orderly relations'), but it does not specify explicit train/validation/test dataset splits for the entire datasets used for model evaluation.
Hardware Specification Yes All the experiments are done using Matlab 2017 in an Intel Core 2.50GHZ desktop.
Software Dependencies Yes All the experiments are done using Matlab 2017 in an Intel Core 2.50GHZ desktop.
Experiment Setup Yes For OSC, we varied the regularization parameters α and β within [0.1, 0.2, 0.3, 0.4, 0.5] and [0.01, 0.02, 0.03, 0.04, 0.05], respectively, and δ was fixed to be 0.002 for all the experiments. To construct orderly relations for OSC, we first randomly selected 10% data for each dataset, and then constructed 30 orderly relations for each selected data as in (Chang et al. 2014). [...] q was set as 5 according to (Gong et al. 2017).