Reservoir characterization by calibration of self‐organized map clusters

Abstract

Kohonen's Self Organizing Feature Maps (SOFM) and other unsupervised clustering methods generate groups based on the identification of various discriminating features. These methods seek an organization in the dataset and form relational organized clusters. However, these clusters may or may not have any physical analogues. A calibration method that relates SOM clusters to physical reality is desirable. This calibration method must define the relationship between the clusters and the observed physical properties; it should also provide an estimate of the validity of the relationships. With the development of a calibrated relationship, the whole dataset can be classified. The principal steps, therefore, are the Three-C's "Clustering, Calibration and Classification".

The Hydrothermal System of Solfatara Crater (Campi Flegrei, Italy) Inferred From Machine Learning Algorithms

“…Among these, self-organizing (or Kohonen) maps (SOM: Kohonen, 2013) are clustering methods based on a competitive learning approach. SOM are regarded as one of the most important tools for unsupervised seismic facies analysis (Taner et al, 2001;Coléou et al, 2003).…”

Seismic facies analysis through musical attributes

“…et al 1994;Fournier and Derain 1995;Johann, Castro and Barroso 2001;Saggaf, Toksöz and Marhoon 2003). Other techniques are support vector machines (Li and Castagna 2004), unsupervised techniques such as self-organising maps (SOMs) (Kohonen 2001;Taner et al 2001;Zhang, Quieren and Schuelke 2001;Coléou, Poupon and Azbel 2003;, 2004a,b, 2007Roy and Marfurt 2011), and, recently, generative topographic mapping (Roy et al 2014;Chopra and Marfurt 2014). In fact, one of the most attractive applications of SOMs in typical problems regarding seismic facies analysis has been performed by Saraswat and Sen (2012).…”

Clinical Decision Support Systems for Comorbidity: Architecture, Algorithms, and Applications

“…Based on the different training schemes, numerous ANN methods are presented in the literature. In the context of wireless health, two main ANN based algorithms are widely employed in a CDSS [ 9 ], including clustering, self-organizing map: clustering based training scheme is implemented by extracting a few critical features of cases and then mapping these features into the structure of an ANN [ 8 ]; self-organizing map alters the structure of an ANN by adding necessary neurons and deleting unnecessary neurons [ 7 ].…”