MGI: an incremental bottom-up algorithm

We discuss an incremental method which learns a concept definition from examples and counter-examples. We present two bottom-up algorithms: MG and MGI that learn characteristic disjunctive definitions; MG is a non-incremental algorithm and MGI is incremental. Some experiments are performed, and a comparison with incremental top-down algorithms shows that MGI is faster, requires fewer instances to learn and found better concept definitions.<
>