Learning Probabilistic Concept Hierarchies

1
LearningProbabilistic Concept Hierarchies

• Cobweb
• / \Arachnid, Twilix, Oxbow

2
Machine learning

• Leer methode
• Supervised
• Unsupervised ?
• Trainings data
• Labeled data
• Unlabeled data ?
• Model
• Tree ?
• Decision rules

3
Cobweb(2)

• Learning method
• Extend downward
• Creating a disjunct
• Merging two categories
• Splitting a concept
• Category Utility function

4
Cobweb

• Cons
• Sensitive to noise
• Sensitive for training order
• Enforces disjunct categories
• No room for sequential representations

5
Arachne

• Tackles noise and training order problem
• How?
• Checks if vertically well placed
• Checks if horizontally well placed

6
Twilix

• Tackles disjunct categories problem
• How?
• By making a tree where the children of each node
  are sets of conflict sets that emphasize
  different attributes.

7
Oxbow

• Tackles sequential representations
• How?
• ??

8
(No Transcript)