Dictionary Learning (DL) is an unsupervised representation learning method in which one can learn the representation (the dictionary) from a set of training data, instead of standard methods that represent the data as a linear combination of basis functions such as Fourier or wavelets. The learning of dictionaries focuses on learning the 'basis' and features by matrix factorization. Recently, this method has been combined with the Deep Learning paradigm obtaining a new representation model called Deep Dictionary Learning (DDL). In this talk, we present an introduction of DL through the task of image denoising. Then, we present two DL multi-model based applications. In the first one, we propose a multimodel DL method for the super-resolution problem. In the second one, we propose a weighted multimodel DDL for tumor segmentation in Magnetic resonance imaging. We also present experiments and preliminary results of both methods.
Registre-se com antecedência para este seminário: https://ide-fgv-br.zoom.us/meeting/register/tJAoc-utpj0vEtAbyi4JyM1v0DZaKhTUs6dd
Após o registro, você receberá um e-mail de confirmação contendo informações sobre conexão no seminário.
*Texto informado pelo autor.
Oscar Dalmau Cedeño received the M.Sc. degree in computer science and industrial mathematics and the Ph.D. degree in computer science from the Mathematics Research Center (CIMAT), Guanajuato, Mexico, in 2004 and 2010, respectively. He is currently researcher and professor of the Department of Computer Science at CIMAT, Mexico. His research interests lie in the areas of machine learning, optimization, image processing, and computer vision.