Published online by Cambridge University Press: 09 March 2009
A learning algorithm for the inference of sequential machines from input/output behaviour is developed. The construction of the models is sequentially achieved by processing input/output (i/o) sequences, and through an induction-contradiction-discrimination scheme. The various states are discriminated after the discovery of contradictory i/o sequences. The properties of the inferred machines and some reported experiments indicate the efficiency of the learning concept of contradiction for this application. Possible application in Robotics for the learning of assembling models is also mentioned.