In Representation and Recognition in Vision (see record 1999-02621-000), Shimon Edelman proposes a solution to the object-recognition problem that challenges the David Marr's view of relating individual representations to individual objects. Edelman's solution to the representation problem (drawing heavily on the work of Roger Shepard) is to represent relationships between objects, rather than representing the objects themselves. This book describes an instantiation of Edelman's theory called the Chorus of Prototypes. Particular views of particular objects are represented explicitly in a multidimensional shape space. These reference objects are termed active landmarks. When novel images are encountered, which will be relatively often, Chorus of Prototypes computes the similarity of the new image to a set of relevant landmarks. At this stage, it is too early to judge Edelman's model. As his model becomes more complex and encodes landmarks in a more biologically plausible fashion, more critical examination of the connection between the model and human data will become possible. In the end, Edelman succeeds in proposing a framework which will provide interest, though not necessarily revolutionary fervor, from those around him.
|Number of pages
|Contemporary Psychology : APA Review of Books
|Published - Jun 2001