A sub-symbolic process underlying the usage-based acquisition of a compositional representation: Results of robotic learning experiments of goal-directed actions

We propose a sub-symbolic connectionist model in which a compositional system self-organizes by learning a provided set of goal-directed actions. This approach is compatible with an idea taken from usage-based accounts of the developmental learning of language. The model explains a possible continuous process underlying the transitions from rote knowledge to systematized knowledge by drawing an analogy to the formation process of a regular geometric arrangement of points. An experiment was performed using a simulated mobile robot reaching or turning toward a colored target. By using an identical learning model, three different types of combinatorial generalization are observed depending on the provided examples. Based on the experimental results, a dynamical systems interpretation of conventional usage-based models is discussed.