The source code for our paper: Symmetrical Self-representation and Data-grouping Strategy for Unsupervised Feature Selection
File SSDS.m is the function of our method to call from Matlab.
MAT_CSV_classification_results_of_our_method_SSDS has 220 mat or csv files in zip. MAT_CSV_clustering_results_of_our_method_SSDS has 440 mat or csv files in zip. FIG_curves_various_number_of_features has 22 fig files.
One can use these files to make a comparison with their new method. The runing result are in mat files. The selected feature index vectors are in csv files. The figure generated by our runing are provided by .fig files, which can be reused by adding new curves to the figure convinently.
File StepsToReproduce.mlx contains the steps to generate the experiment results, and all needed support function. The dataset mentioned in the code can be found in the following link: https://jundongl.github.io/scikit-feature/datasets.html
- Put File StepsToReproduce.mlx to the working directory of Matlab.
- Put the data of .mat files to the sub-folder of './datasets/'. Change accordingly the dataset file name in the variable of "datalist";
- Change the value of parameters or the range of parameters, if needed. (In the 1st code section) The parameters are alpha, beta, gamma, sigma and the number of groups (group_num)
- Press Run button to have a look at the result output. This main file test all parameters from the pre-defined ranges one by one.