| 研究実績の概要 |
We proposed Bayesian method for Tucker decomposition and Tucker completion. By employing group sparsity priors over factor matrices and cores, our method can automatically learn the multilinear rank from given observed tensor data.
We developed tensor denoising algorithm based on Bayesian low-rank tensor factorization, which can be used for image, video and MRI denoising. Furthermore, we apply Bayesian tensor methods to EEG signal analysis and human action recognition from video data.
|
