Searching Version Spaces

Publication	Mitchell/97b: Machine Learning
Name	Searching Version Spaces
Description	If we can manage to find an appropriate ordering of our hypothesis language, then learning can be regarded as navigating through version spaces, with each new example read limiting the number of hypotheses, that are still consistent. The search will either lead from specific to general, driven by positive examples (bottom-up), or from general to specific, driven by negative examples (top-down). The CANDIDATE-ELIMINATION Algorithm combines a bottom-up and a top-down search.
Specialization	CANDIDATE-ELIMINATION
Hypothesis Language	Version Space
Dm Step	Concept Learning
Method Type	Method