Browsing by Author "Guivarch, Ronan"
Now showing 1 - 5 of 5
Results Per Page
Sort Options
- An hybrid approach for the parallelization of a block iterative algorithmPublication . Balsa, Carlos; Guivarch, Ronan; Ruiz, Daniel; Zenadi, MohamedThe Cimmino method is a row projection method in which the original linear system is divided into subsystems. At every iteration, it computes one projection per subsystem and uses these projections to construct an approximation to the solution of the linear system. The usual parallelization strategy in block algorithms is to distribute the different blocks on the available processors. In this paper, we follow another approach where we do not perform explicitly this block distribution to processors within the code, but let the multi-frontal sparse solver MUMPS handle the data distribution and parallelism. The data coming from the subsystems defined by the block partition in the Block Cimmino method are gathered in an unique block diagonal sparse matrix which is analysed, distributed and factorized in parallel by MUMPS. Our target is to define a methodology for parallelism based only on the functionalities provided by general sparse solver libraries and how efficient this way of doing can be.
- An hybrid approach for the parallelization of a block iterative algorithmPublication . Balsa, Carlos; Guivarch, Ronan; Ruiz, Daniel; Zinadi, MohamedThe Cimmino method is a row projection method in which the original linear system is divided into subsystems. At every iteration, it computes one projection per subsystem and uses these projections to construct an approximation to the solution of the linear system. The usual parallelization strategy in block algorithms is to distribute the different blocks on the available processors. In this paper, we follow another approach where we do not perform explicitly this block distribution to processors within the code, but let the multi-frontal sparse solver MUMPS handle the data distribution and parallelism. The data coming from the subsystems defined by the block partition in the Block Cimmino method are gathered in an unique block diagonal sparse matrix which is analysed, distributed and factorized in parallel by MUMPS. Our target is to define a methodology for parallelism based only on the functionalities provided by general sparse solver libraries and how efficient this way of doing can be
- Dental image segmentation by clustering methodsPublication . Balsa, Carlos; Alves, Cláudio; Guivarch, Ronan; Mouysset, SandrineSegmentation of dental radiography allows the identification of human individuals but also could be used for the development of more effective diagnostic, monitoring, and evaluation of appropriate treatment plans. In practice, dark background and bones tissues are not distinguished with contour extraction methods on dental images. So we propose to first apply the k-means method and then extract the contours on the clustering result. We present an initialization of the k centroids based on the grey scale histograms, a weighted norm that includes both grey scale and geometrical information, and tests it on dental X-ray images. Then we describe a promising parallel clustering method based on kernel affinity.
- MUMPS based approach to parallelize the block cimmino algorithmPublication . Balsa, Carlos; Guivarch, Ronan; Raimundo, João; Ruiz, DanielThe Cimmino method is a row projection method in which the original linear system is divided into subsystems. At every iteration, it computes one projection per subsystem and uses these projections to construct an approximation to the solution of the linear system. The usual parallelization strategy applied in block algorithms is to distribute the different blocks on the different available processors. In this paper, we follow another approach where we do not perform explicitely this block distribution to processors whithin the code, but let the multi-frontal sparse solver MUMPS handle the data distribution and parallelism. The data coming from the subsystems defined by the block partition in the Block Cimmino method are gathered in an unique matrix which is analysed, distributed and factorized in parallel by MUMPS. Our target is to define a methodology for parallelism based only on the functionalities provided by general sparse solver libraries and how efficient this way of doing can be. We relate the development of this new approach from an existing code written in Fortran 77 to the MUMPS-embedded version. The results of the ongoing numerical experiments will be presented in the conference
- Optimal Latent Variables Number for the Reconstruction of Time Series with PLSRPublication . Balsa, Carlos; Dupuis, Hugo; Breve, Murilo Montanini; Guivarch, Ronan; Rufino, JoséThe Partial Least Squares Regression is an efficient method for the filling of gaps in meteorological time series. It enables to reduce the dimension of the predictor dataset to a reduced number of latent variables, without loss of significant information. Defining the number of latent variables to be used is an essential aspect of the success of the method. This study is about the comparison between eight different criteria, used in the choice of latent variables. The results indicate that the criteria based on cross-validation are the most efficient, being, however, more computationally demanding.