Aaaaa
Páginas: 2 (254 palabras)
Publicado: 19 de noviembre de 2012
Resumen Paper IA II
- The main objective of self-organizing maps is data clustering and their graphical presentation .
- A comparative analysishas been made for two data sets: Fisher’s iris data set
and the economic indices of the European Union countries .
- A new SOM system is alsointroduced and researched.
- It helps to see the proportion of neurons, corresponding to the data items, belonging to the different classes, and fallenin the same SOM cell.
- As the target of the SOM is data clustering and visualization, the systems are assessed according to their ability to clusterand visualize data correctly -> creo que se va a evaluar cada tecnica en base a estos dos items.
4. A new system of SOM
However systems havea common disadvantage. If the classes, to which the data belong to, are known, and the labels of the classes are displayed in the map, it is difficultto understand how much the vectors from one or other class correspond to a cell (neuron), because usually only different (but not the same) labels areshown .
-> Creo que este nuevo sistema de SOM se basa en la idea de ver si los datos que han activado una neurona pertenecen a esa clase, para estoantes de entrenar cada neurona del mapa debe haber sido etiquetada con un etiqueta (supongo que esta etiqueta ya se conocia).
-> Problema que seintenta resolver: saber cuantos datos y de que clases activaron una neurona.
-> IMPORTANTE : usa una nueva funcion de vecindad que esta media fea.
- The main objective of self-organizing maps is data clustering and their graphical presentation .
- A comparative analysishas been made for two data sets: Fisher’s iris data set
and the economic indices of the European Union countries .
- A new SOM system is alsointroduced and researched.
- It helps to see the proportion of neurons, corresponding to the data items, belonging to the different classes, and fallenin the same SOM cell.
- As the target of the SOM is data clustering and visualization, the systems are assessed according to their ability to clusterand visualize data correctly -> creo que se va a evaluar cada tecnica en base a estos dos items.
4. A new system of SOM
However systems havea common disadvantage. If the classes, to which the data belong to, are known, and the labels of the classes are displayed in the map, it is difficultto understand how much the vectors from one or other class correspond to a cell (neuron), because usually only different (but not the same) labels areshown .
-> Creo que este nuevo sistema de SOM se basa en la idea de ver si los datos que han activado una neurona pertenecen a esa clase, para estoantes de entrenar cada neurona del mapa debe haber sido etiquetada con un etiqueta (supongo que esta etiqueta ya se conocia).
-> Problema que seintenta resolver: saber cuantos datos y de que clases activaron una neurona.
-> IMPORTANTE : usa una nueva funcion de vecindad que esta media fea.
Leer documento completo
Regístrate para leer el documento completo.