calcumat | Compute a U-matrix |
-s <som> | name of the TS-SOM |
-d <data> | data for distance computing |
-umat <umat> | name for the resulting U-matrix data |
grpumat | U-matrix for groups |
-s <som> | name of the SOM |
-w <weights> | prototype data |
-c <groups> | classified data for the groups |
-fout <field-out> | resulting field |
[-in] | within groups, default is between |
This command computes a U-matrix for a TS-SOM. <data> should contain those variables of the neurons, which are used for computing the distances between neurons. The result is a data frame <umat> containing the distance information between neurons. The operation creates a data frame that contains one data record for every neuron. The distances between neurons are stored in fields (2 directions SOM dimensions). The data frame will be <umat>.data. In addition, the operation stores the minimum and maximum values in each layer in frame <umat>.dminmax. For instructions, how to visualize the U-matrix, see umat (section 7.2.6).
grpumat collects U-matrix distances into one field based on specified groups. If flag -in is not given, then distances between neurons of distinct groups are collected, whereas with the flag only distances between the neurons belonging to the same group are picked.
Example: As an example, three weights rm_w, zn_w and age_w are selected to be the basis of distance computing. Then, a U-matrix is computed and created into a graphic structure. To see the result, run the example `ex7.10' (section 7.2.6).
... NDA> somtr -d boston -sout s1 -l 5 ... NDA> select umatflds -f s1_W.rm_w s1_W.zn_w s1_W.age_w NDA> calcumat -s s1 -d umatflds -umat umat1 NDA> ls -fr umat1 umat1.data umat1.data.d_0_0 umat1.data.d_0_1 umat1.data.d_1_0 umat1.data.d_1_1 umat1.dminmax umat1.dminmax.min umat1.dminmax.max # Set graphical presentation NDA> mkgrp grp1 -s s1 NDA> nsize grp1 -v 1.0 NDA> layer grp1 -l 4 NDA> umat grp1 -umat umat1 ...