How choose matrix size for u-matrix visualization in self organising map of data

[Luna]

Could you please provide a detailed explanation and sample example ?

[AS]

I have a data with dimension 1135 by 30, where each oh row representing the amplitude distribution at aparticular depth. That means for depth 0.5m there is an amplitude distribution of a signal with matrix dimension 1 by 30. Similarly, for other depths also, upto 1135 by 30. Now want to apply Self Organising map (SOM) on this data to classify the classes present in the data. In SOM algorithm there is an concept of U-matrix which is called unified distance matrix, which is claculating the the distance between two data ponits. U matrix is represented by considering appropiate number of msize or munit matrix. I want to know selection of U-matrix for proper classification and also how to interpret this U-matrix?

Answer 1

[AS]

I have a data with dimension 1135 by 30, where each oh row representing the amplitude distribution at aparticular depth. That means for depth 0.5m there is an amplitude distribution of a signal with matrix dimension 1 by 30. Similarly, for other depths also, upto 1135 by 30. Now want to apply Self Organising map (SOM) on this data to classify the classes present in the data. In SOM algorithm there is an concept of U-matrix which is called unified distance matrix, which is claculating the the distance between two data ponits. U matrix is represented by considering appropiate number of msize or munit matrix. I want to know selection of U-matrix for proper classification and also how to interpret this U-matrix?