Multimodal graphical models via Group Lasso
Resumen:
Graphical models are a very useful tool to describe and understand natural phenomena, from gene expression and brain networks to climate change and social interactions. In many cases, the data is multimodal. For example, one may want to build one network from several fMRI (functional magnetic resonance imaging) studies from different subjects, or combine different data modalities (as fMRI and questions) for several subjects. To this end, in this work we combine group lasso with graphical lasso, and derive an iterative shrinkage thresholding algorithm for solving the proposed optimization problem. The framework is validated with synthetic data and real fMRI data, showing the advantages of combining different modalities in order to infer the underlying network structure.
| 2013 | |
| Procesamiento de Señales | |
| Inglés | |
| Universidad de la República | |
| COLIBRI | |
| https://hdl.handle.net/20.500.12008/41761 | |
| Acceso abierto | |
| Licencia Creative Commons Atribución - No Comercial - Sin Derivadas (CC - By-NC-ND 4.0) |
| Sumario: | Graphical models are a very useful tool to describe and understand natural phenomena, from gene expression and brain networks to climate change and social interactions. In many cases, the data is multimodal. For example, one may want to build one network from several fMRI (functional magnetic resonance imaging) studies from different subjects, or combine different data modalities (as fMRI and questions) for several subjects. To this end, in this work we combine group lasso with graphical lasso, and derive an iterative shrinkage thresholding algorithm for solving the proposed optimization problem. The framework is validated with synthetic data and real fMRI data, showing the advantages of combining different modalities in order to infer the underlying network structure. |
|---|
