run for X = 50,100,150,200,250,300,350,400,all (so all graphs
get 9 x values) and send me the results in exactly the same format

and try the following variants:

1) {log,noLog} the attributes

2) it is known tht descritization can improve naiveBayes. so, for
NB, try {none,3bins,5bins,10bins} , equal width bins (and for none,
use the current gaussian)

note that combinations of (1) * (2) will include logging and not logging , then descritizing, not descritizing.

for all the above, generate graphs like the pdfs in results/*.pdf