Learning and recognition of sketches for complex computational cognition - Position paper

Abstract

In order to enable machines to operate intelligently in their environment, it is important that they do not only collect sensory input about their environment, but also recognize and understand objects. Analogical reasoning is considered fundamental for many complex cognitive processes. In this paper, we present an experiment which gives empirical support of our hypothesis that object recognition and concept formation rely fundamentally on analogical similarities. Similar object sketches with the same structure are recognized faster and more frequently than similar object sketches with different structure. Afterwards, we introduce our analogy-making framework Heuristic-Driven Theory Projection (HDTP) and explain how HDTP can be used for object recognition.