Compression

Craw = (?i ? 1..n N(Hi)) - (n-1)*N(Hp)

i.e. sum of hbonds in the examples minus (No_of_examples - 1) * hbonds in

pattern

(2) Normalised compression:

Cnorm = 1 - ((?i ? 1..n N(Hi))- Craw) / . ((?i ? 1..n N(Hi))- min(N(Hi)))

This is a goodness of compression measure (0=good, 1=bad).