Structure Regularized Unsupervised Discriminant Feature Analysis

Authors: Mingyu Fan, Xiaojun Chang, Dacheng Tao

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Clustering and classification experiments on real world image data sets demonstrate the effectiveness of our method.
Researcher Affiliation Academia Mingyu Fan College of Maths & Info. Science, Wenzhou University, Wenzhou 325035, China fanmingyu@wzu.edu.cn Xiaojun Chang Centre for Artificial Intelligence University of Technology Sydney, Sydney, NSW 2007, Australia cxj273@gmail.com Dacheng Tao Centre for Artificial Intelligence University of Technology Sydney, Sydney, NSW 2007, Australia dacheng.tao@uts.edu.au
Pseudocode Yes Algorithm 1 ADMM for solving problem (8) and Algorithm 2 The algorithm for solving problem (13)
Open Source Code No The paper does not provide an explicit statement or a link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes The experiments are conducted on publicly available image data sets: the Coil-20 data set (Coil-20)1, the USPS handwritten digits data set (USPS)2, the Yale-B Extended (Yale B) and the CMUPIE face data sets3. 1http://www.cs.columbia.edu/CAVE/software/softlib/coil20.php 2http://www.escience.cn/people/fpnie/ 3http://www.cad.zju.edu.cn/home/dengcai/Data/Face Data.html
Dataset Splits Yes Given a data set, we randomly select p percents from every class in data X to formulate the training set Xtrain, and the left data are used as the test data Xtest. The results when p = 30 is presented here and the results when p = 10 is provided in the supplement material.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup Yes For methods require neighborhood sizes for data structure learning, the neighborhood size is searched in {4, 6, 8, 10}. For UDFS, RUFS, and JELSR, the regularization parameters are searched in the range {10 5, 10 4, , 101, 102}. The parameter γA is searched in the range {0.01, 0.05, 0.1, 0.5, 1}. To make the experimental results reproducible, λZ for Sr-UDFS is set as 0.1 respectively throughout the experiments.