Using Riemannian Geometry for Source Separation

Finding hidden variables in data can be challenging. Using Blind Source Separation allows to exact the independent sources in a unsupervised way. However, the methods require usually strong optimization constraints in order to avoid degenerate solutions and/or improve convergence. By using Riemannian Geometry framework, we propose to use the geometrical properties of the manifold that […]

EUSIPCO16: Presenting Composite Approximate Joint Diagonalization

We were in EUSIPCO 2016 conference in Budapest to present our work on Blind Source Separation with Approximate Joint Diagonalization. We shared the idea that using several data models simultaneously can improve the robustness of the data mining. Our composite model was used to discriminate Event-Related Potential sources from background brain activity with low resolution […]