MATLAB implementation of "Nearly Optimal Robust Subspace Tracking", ICML 2018. Longer version to appear in IEEE Journal of Selected Areas in Information Theory, 2020.
We use self-expressive layer to learn the affinity matrix of the hidden space of a generator, and we found subspace in GAN. Also we propsed a subspace-based high-fidelity GAN inversion model.
We propose a global and local feature transformation method for PRID. The global feature transformation matrix projects the data from different cameras to a common space. We further hypothesize that a latent basis matrix can be learnt in this space which represents the shared structure between different cameras using matrix factorization.
[IEEE TSP 2021] “Robust Subspace Tracking with Missing Data and Outliers: Novel Algorithm with Convergence Guarantee”. IEEE Transactions on Signal Processing, 2021.