CAN A LINEAR AUTOASSOCIATOR RECOGNIZE FACES FROM NEW ORIENTATIONS

An often noted limitation of computational models of faces operating o n two-dimensional pixel intensity representations is that they cannot handle changes in orientation. We show that this limitation can be ove rcome by the use of multiple views of a given face instead of a single view to represent the face. Specifically, we show that a linear autoa ssociator trained to reconstruct multiple views of a set of faces is a ble to recognize the faces from new view angles. An analysis of the in ternal representation of the memory (i.e., eigenvectors of the between -unit-connection weight matrix) shows a dissociation between two kinds of perceptual information: orientation and identity information. (C) 1996 Optical Society of America