Unsupervised Learning of Visual Structure Ben Hiles Cornell University bph7@cornell.edu   How object and scene structure is represented is an important problem in vision. Considerations of coding efficiency suggest that representations of structurally related objects should share common fragments. Can a useful set of object fragments be acquired in an unsupervised fashion? One possible solution to this problem is the use of statistical learning. Statistical interdependence criteria such as pairwise conditional probabilities of the fragments, or Barlow's "suspicious coincidence" ratio (the joint probability of two fragments divided by the product of their marginal probabilities), are two possible measures. If humans use such criteria in learning structured objects, they would tend to lump together a pair of highly interdependent fragments, perceiving them as a single shape. This hypothesis was tested in an experiments involving a same/different task. As predicted, exposure to a set of statistically controlled stimuli led to a shorter reaction time for fragments with higher interdependence (conditional probability). This learning was done implicitly, since none of the subjects reported noticing the underlying statistics. This study complements and extends recent results by Aslin and others that demonstrate some of the learning strategies used by the brain in dealing with structured stimuli. This work was supervised by Shimon Edelman.   Back to main page