machine-learning learning-theory
Definition
Consistent Hypothesis
A hypothesis is said to be consistent with a training dataset if it correctly classifies every instance in the sample. Formally:
Equivalently, a consistent hypothesis is one that achieves a zero empirical risk: .
Role in Learning Theory
In the PAC learning framework under the realisability assumption, the learning algorithm is guaranteed to find at least one consistent hypothesis. The objective of the theory is to determine the conditions (sample complexity) under which any hypothesis consistent with the training set is also probably approximately correct on the full distribution.