Estimation Under Model-Based Sparsity

摘要: Beyond the ordinary, extensively studied, plain sparsity model, a variety of structured models have been proposed in literature Bach (2008); Roth and Fischer Jacob et al. (2009); Baraniuk (2010); (2012); Chandrasekaran Kyrillidis Cevher (2012a). These are designed to capture interdependence locations non-zero components that is known priori certain applications. For instance, wavelet transform natural images often (nearly) sparse dependence among dominant coefficients can be represented by rooted connected tree. Furthermore, applications such as array processing or sensor networks, while different sensors may take measurements, support set observed signal identical across sensors. Therefore, model this property system, we compose an enlarged with jointly-sparse block-sparse set, whose occur contiguous blocks.