/DSRL

Primary LanguagePython

Deep Sparse Regularizer Learning (DSRL)

This is an implement of methods in "Learning Deep Sparse Regularizers with Applications to Multi-View Clustering and Semi-Supervised Classification" that published in IEEE Trans. Pattern Analysis and Machine Intelligence.

Datasets Descriptions

  • Original similarity matrices are stored in /datasetW, which are generated by KNN (see Matlab codes in ConstructW.zip).
  • Original multi-view datasets are stored in /_multiview datasets.

Environment

Require Python 3.8

  • torch 1.8.0
  • numpy 1.16.3
  • tqdm 4.28.1
  • scikit-learn 0.20.3

Quick Running

  • Run python ./run_Clustering.py --dataset-id 1 for clustering tasks.
  • Run python ./run_Classification.py --dataset-id 1 for semi-supervised classification tasks.
  • Note: dataset-id values of all presented datasets are as shown below:
    • 1:'ALOI', 2:'Caltech101-7', 3:'Caltech101-20', 4:'MNIST', 5:'MSRC-v1', 6:'NUS-WIDE', 7:'Youtube', 8:'ORL'

Citation

  • Shiping Wang, Zhaoliang Chen, Shide Du, and Zhouchen Lin, Learning Deep Sparse Regularizers with Applications to Multi-View Clustering and Semi-Supervised Classification, IEEE Trans. Pattern Analysis and Machine Intelligence, 2021.