Gene clustering: 2001:
2
Properties of Domain
§
Small population of instances
(1000's)
•
vs
small training set from large population
§
All instances have known class
•
vs
only training set classified
§
Large descriptions of each instance
•
common in text and biological domains
§
Need characteristic
,
not discriminant
,
models
•
human understanding is the goal,
•
not automatic classification of future instances