Matlab source codes of the Probabilistic Rank-One Discriminant Analysis (PRODA) algorithm presented in the paper Probabilistic Rank-One Discriminant Analysis via Collective and Individual Variation Modeling.

Face recognition with PRODA on 2D images from the FERET dataset:

Demo_PRODA.m

• DBpart.mat stores the indices for training (2 samples per class)/test data partition.
• FERETC80A45.mat stores 320 faces (32x32) of 80 subjects (4 samples per class) from the FERET dataset.
• Demo_PRODA.m provides example usage of PRODA for subspace learning and classification on 2D facial images.
• PRODA.m implements the PRODA algorithm described in the paper.
• projPRODA.m projects 2D data into the subspace learned by PRODA.
• sortProj.m sorts features by their Fisher scores in descending order.
• logdet.m computes the logarithm of determinant of a matrix.

Tensor toolbox v2.6.

If you find our codes helpful, please consider cite the following paper:

@article{
    zhou2020PRODA,
    title={Probabilistic Rank-One Discriminant Analysis via Collective and Individual Variation Modeling},
    author={Yang Zhou and Yiu-ming Cheung},
    journal={IEEE Transactions on Cybernetics},
    year={2020},
    volume={50},
    number={2},
    pages={627-639},
    doi={10.1109/TCYB.2018.2870440},
}