Next:
Up:
Previous:
- Direct Feedback: Evaluates each move (quality of board)
- Indirect Feedback: Only knows final outcome (win, lose, or draw),
must deal with credit assignment problem
- Supervised Training: Depends on teacher
- Query-based Training: May pose examples to teacher for evaluation
- Exploration: May ``play'' itself (no teacher)
- Unsupervised: Clustering (finding groups of ``simmilar'' objects) or
Associations (finding that some values of attributes in a database appear with
others)
A problem: is training experience representative of performance goal?
