Version SpaceIt is the subset of all hypotheses in concept learning that are consistent with the known dataset, and is usually used to converge on content.

Version space learning is a logical approach to machine learning, specifically binary classification. Version space learning algorithms search a predetermined space of hypotheses, viewed as a set of logical statements.

For the "rectangular" hypothesis in two-dimensional space (right figure), the green plus sign represents the positive sample, and the red circle represents the negative sample. GB is the maximally generalized positive hypothesis boundary, and SB is the maximally specific positive hypothesis boundary. The rectangle in the area enclosed by GB and SB is the hypothesis in the version space, that is, the area enclosed by GB and SB is the version space.

In some cases where the generalization ability of hypotheses needs to be ranked, the version space can be represented by the two upper and lower bounds GB and SB. During the learning process, the learning algorithm can only operate on the two representative sets GB and SB.

References

【1】https://www.jishux.com/p/1eaaad466795eb5c