smhsom | Smooth data vectors in neurons |
-s <som> | name of the TS-SOM structure |
-d <data> | data to be smoothed |
-dout <dataout> | name of the resulting data |
[-hits <hitfield>] | field including hitcounts for clusters |
[-max <maxit>] | maximum number of iter. (default 50) |
This operation smooths a given data according to the structure of a SOM. Data records corresponding to neurons are averaged by the values of their neighbors. If <hitfield> has been specified, then only empty neurons are recomputed; otherwise updating is performed for all the neurons.
Example (ex5.3): This example shows the effect of smoothing.
... NDA> somtr -d pre -sout som -cout cld -l 5 NDA> clstat -d boston -c cld -dout sta -all NDA> smhsom -s som -d sta -dout smosta -hits sta.hits -max 20 NDA> mkgrp grp1 -s som NDA> setgdat grp1 -d sta NDA> layer grp1 -l 4 NDA> bar grp1 -f sta.crim_avg -co red -sca som NDA> bar grp1 -f sta.chas_avg -co blue -sca som NDA> mkgrp grp2 -s som NDA> setgdat grp2 -d smosta NDA> layer grp2 -l 4 NDA> bar grp2 -f smosta.crim_avg -co red -sca som NDA> bar grp2 -f smosta.chas_avg -co blue -sca som
Original Smoothed