Properties of Candidate-Elim. Algo.

• independent of the order in which the training data is presented
• S and G boundaries move monotonically closer to each other
• will converge if
  1. there are no errors in the training examples
  2. there is some hypothesis in that correctly describes the target concept
• can determine when sufficient training examples have been observed to converge, and are identical
• can detect errors or bad representation by convergence to the empty version space