DP-space: Bayesian Nonparametric Subspace Clustering with Small-variance Asymptotics

Authors: Yining Wang, Jun Zhu

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate that DP-space outperforms various competitors in terms of clustering accuracy and at the same time it is highly efficient. We compare DP-space with several competitors on both synthetic and real-world datasets.
Researcher Affiliation Academia Yining Wang YININGWA@CS.CMU.EDU Machine Learning Department, Carnegie Mellon University, Pittsburgh, PA 15213, USA Jun Zhu DCSZJ@TSINGHUA.EDU.CN Dept. of Comp. Sci. & Tech., State Key Lab of Intell. Tech. & Sys., TNList, CBICR Center, Tsinghua University, China
Pseudocode Yes Algorithm 1 The DP-space algorithm
Open Source Code No The paper refers to code for *other* subspace clustering algorithms and implementations used for comparison, but does not provide access to the source code for their own DP-space method.
Open Datasets Yes We compare DP-space with several competitors on both synthetic and real-world datasets. The dataset contains 10, 000 data points from R3. Table 2. Average classification error (%) and running time (seconds) on the Hopkins-155 dataset.
Dataset Splits Yes Parameters of the DP-space algorithm are selected using 30% of the ground-truth labels according to NMI. The other hyper-parameters are selected via cross-validation.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments.
Software Dependencies No The paper states 'All methods are implemented in Matlab', but does not provide specific version numbers for Matlab or any other software libraries used.
Experiment Setup Yes Throughout the experiments we always set the hyper-parameter a to be 1, under which no shrinkage on lj is imposed. Values of λ range from 10 3 to 102 and values of s range from 10 2 to 102. We run the EM-MPPCA algorithm using 10 random initializations on each video sequence.