Incremental Learning with Ordinal Bounded Example Memory

A Bounded Example Memory learner is a learner that operates incrementally and maintains a memory of finitely many data items. The paradigm is well-studied and known to coincide with setdriven learning. A hierarchy of stronger and stronger learning criteria is obtained when one considers, for each k ∈ N, iterative learners that can maintain a memory of at most k previously processed data items. We report on recent investigations of extensions of the Bounded Example Memory model where a constructive ordinal notation is used to bound the number of times the learner can ask for proper global memory extensions.